Investigating the Combination of Bag of Words and Named Entities Approach in Tracking and Detection Tasks among Journalists
- Author: Mohd Masnizah, Bashaddadh Omar Mabrook A.
- Publish: Journal of Information Science Theory and Practice Volume 2, Issue4, p31~48, 31 Dec 2014
The proliferation of many interactive Topic Detection and Tracking (
iTDT) systems has motivated researchers to design systems that can track and detect news better. iTDT focuses on user interaction, user evaluation, and user interfaces. Recently, increasing effort has been devoted to user interfaces to improve TDT systems by investigating not just the user interaction aspect but also user and task oriented evaluation. This study investigates the combination of the bag of words and named entities approaches implemented in the iTDT interface, called Interactive Event Tracking ( iEvent), including what TDT tasks these approaches facilitate. iEvent is composed of three components, which are Cluster View (CV), Document View (DV), and Term View (TV). User experiments have been carried out amongst journalists to compare three settings of iEvent: Setup 1 and Setup 2 (baseline setups), and Setup 3 (experimental setup). Setup 1 used bag of words and Setup 2 used named entities, while Setup 3 used a combination of bag of words and named entities. Journalists were asked to perform TDT tasks: Tracking and Detection. Findings revealed that the combination of bag of words and named entities approaches generally facilitated the journalists to perform well in the TDT tasks. This study has confirmed that the combination approach in iTDT is useful and enhanced the effectiveness of users' performance in performing the TDT tasks. It gives suggestions on the features with their approaches which facilitated the journalists in performing the TDT tasks.
Bag of Words , Named Entity Recognition , Interactive Topic Detection and Tracking , User , Journalists
A vast amount of information arrives every day through a variety of news media such as newswires, TV, newspapers, radio, and website sources. In addition, the significant growth and dynamic environment of digital information creates challenges in information retrieval (IR) technology. The amount of constant information which readers can effectively use is still limited compared to the continuous growth of rapidly changing online information. Thus it has become increasingly important to provide and enhance methods to find and present textual information effectively and efficiently. Technologies have been proposed to solve the information overload issues, including information customization, search engines, and information agency. Therefore research on Topic Detection and Tracking (TDT) serves as a core technology for a system that monitors news broadcasts. It helps to alert information professionals such as journalists to interesting and new events happening in the world. Journalists are keen to track and, in particular, to know the latest news about a story from a large amount of information that arrives daily.
Most TDT research has concentrated primarily on the design and evaluation of algorithms to implement TDT tasks such as stream segmentation, story detection, link detection, and story tracking. Bag of words (BOW) and named entities (NE) approaches are widely implemented either in document representation or in user interface design of TDT systems (Kumaran & Allan, 2004). Meanwhile Interactive Topic Detection and Tracking (
iTDT) refers to the TDT works which focus on user interaction, user evaluation, and user interfaces aspects (Mohd et al., 2012).
In this research we did not discuss the interface design or features but focused on the approaches used: either BOW or NE or their combination. The objectives of this research are: (i) to implement the combined approach of BOW and NE in
iTDT interface and, (ii) to investigate the usability of this approach on the iTDT interface in assisting journalists to perform the TDT tasks. This is important to answer questions, e.g., which approach aims to facilitate journalists to perform TDT tasks? Will the use of a combined approach of BOW and NE improve iTDT? What are the TDT tasks facilitated through the use of this combined approach?
The focus of this section is to review related works on
iTDT with a view to identifying best approaches used in the design of iTDT interfaces. The previous scholarly studies in TDT tried to build better document models, enhancing and developing similarity metrics or document representations. A large number of research efforts have focused on enhancing and improving document representations by applying Natural Language Processing (NLP) technology such as Named Entity Recognition (NER) (Yang et al., 1999, Makkonen et al., 2004, Kumaran & Allan, 2004, Zhang et al., 2004). Some research tended towards interactive TDT through casting its attention on the graphical user interface (GUI), rather than on laboratory experiments such as TDT Lighthouse (Leuski & Allan, 2000), TimeMine (Swan & Allan, 2000), Topic Tracking Visualization Tool (Jones & Gabb, 2002), Event Organizer (Allan et al., 2005), Stories in Time (Berendt & Subasic, 2009) and Interactive Event Tracking System ( iEvent) interface (Mohd et al., 2012), which are all examples of TDT studies that investigate some approaches to enhancing and improving TDT systems’ and users’ performance. Based on the works reviewed, none of them investigate the effectiveness of their interfaces using the combination of a bag of words (BOW) and named entities (NE) approach. We review and discuss the approaches used in these related studies and we concentrate in iEvent as the baseline setup.
The Interactive Event Tracking System (
iEvent) interface (Mohd et al., 2012) is a prototype system that includes the design of a novel iTDT system interface to facilitate professionals such as journalists and information analysts in performing TDT tasks. This system consists of two settings: Setup 1 is the baseline Setup that uses the BOW approach and Setup 2 is the experimental Setup that uses the NE approach to present the information to the user in a meaningful way using the following components: Cluster View (CV), Document View (DV), and Term View (TV). iEvent incorporates a number of components and features into a single interface, such as cluster labelling, top terms, histogram with the timeline, document content, and keyword. Moreover, it implements good approaches, such as BOW and NE. However, these approaches are only implemented separately, each one in a single setup. Clearly the BOW offers a better explanation in the Detection task as it resembles the What. Meanwhile NE provides specific information on the Who, Where, and When in performing the Tracking task (Mohd et al., 2012). The major problem with this kind of application and approach is that neither supports both tasks (Detection and Tracking) to gather information. So this research attempts to solve this problem by adding a new Setup called Setup 3. This new Setup, which will be considered as an experimental Setup which uses a combination approach (BOW+NE), is intended to offer a better explanation and provide information on the Who, Where, When, and Whatto the user.
The reviewed works in
iTDT used different approaches to extract, organize, and represent information from the news sources to the user such as BOW, NE, query, graphical visualization, text summarization, and other approaches. But none of these works supports tracking and detection tasks at the same time. Therefore, the combination of BOW and NE could enhance user performance in TDT tasks.
There are four features out of seven of
iEvent which distinguish between three approaches: BOW, NE, and a combination of BOW and NE across setups, as apparent in Table 1. The descriptions are:
a. BOW are the keywords aiming to provide the What and were implemented in Setup 1. Most frequent keywords were offered in the features of iEvent. b. NE are the significant keywords that aim to provide the Who, Where and When and were implemented in Setup 2. Most frequent named entities were offered in the features of iEvent. c. Combination of BOW and NE that aim to provide the What, Who, Where and When and were implemented in Setup 3. Most frequent keywords and named entities were offered in the features of iEvent.
The four features of
iEvent that are different between the setups are: ‘CV: cluster labelling,’ ‘CV: top terms,’ ‘DV: document content,’ and ‘TV: keyword approach.’ Each was using a combination of BOW and NE in Setup 3 (experimental setup) and using BOW alone in Setup 1 (baseline setup) and NE alone in Setup 2 (baseline setup). We extracted the named entities using ANNIE (A Nearly-New Information Extraction system), which is an information extraction component of the General Architecture for Text Engineering (GATE). We used it for its accurate entity, pronoun, and nominal co-references extraction (Cunningham et al., 2002).
The Cluster View visualized information related to the size and density of a cluster, and displayed the ten most frequent terms (Top terms) in the cluster. The main difference here is in the Cluster labelling feature (refer to Fig. 1) and top ten terms feature (refer to Fig. 2). Clusters were generated using single pass clustering (Mohd et al., 2012) and were labelled using the three most frequently named entities; for example, Cluster 30 was labelled using keywords in Setup 1, named entities in Setup 2, and a combination of BOW and NE in Setup 3.
When the user clicked on Cluster 30, there was additional information presented on the top ten terms as shown in Fig. 2. The difference in these features between the three setups is that the user was provided with mixed BOW and NE in Setup 3 instead of using BOW only in Setup 1 and NE only in Setup 2.
Document View provided the user with information about the document timeline in a histogram form and the document content in a cluster. There was no difference in timeline across setups where this feature was showing the occurrence and the number of documents (document frequency) for a specific date. This supported the journalist in viewing the information flow in a press article or in analyzing the discourse.
However, as shown in Fig. 3, users could see a small difference in the document content, especially between Setup 3 (experimental Setup) and Setup 2 (baseline Setup) and a significant difference with Setup 1 (baseline Setup). The types of term such as Who, What, Where, and When are highlighted according to the setup.
Cascading Style Sheets (CSS) were used to differentiate three types of named entities with different colors and terms assigned, as shown in Table 2.
Term View provided information on the occurrence of keywords and NE within the timeline with a histogram in a cluster, and an option terms list where there is no difference in timeline across setups. The difference only appeared in the option terms list, where Setup 1 (baseline setup) provided the user with “What” only and Setup 2 (baseline setup) provided the user with “Who, Where, and When”; meanwhile Setup 3 (experimental setup) provided the user with “Who, Where, When, and What.”
The three components were arranged on the interface from top to bottom based on their relevance to the user, beginning with Cluster View (CV), followed by Document View (DV), and lastly Term View (TV). CV enabled the users to browse the whole collection before they went on to search a specific cluster, while DV allowed them to view all the documents in the chosen cluster and provided them with an interactive graphical timeline interface to browse the major events and the cluster contents. Finally, TV displayed the terms in the cluster and provided a timeline to link the terms with the related document in DV. The components were arranged in an inverted pyramid on the interface to help users to narrow their browsing and to be more focused in their searching. In this study,
iEvent had three setups. Setup 1 and Setup 2 (Mohd et al., 2012) are the baseline setups and Setup 3 (Fig. 4) is the experimental setup. iEvent Setup 3 provided users with the same amount of data as in Setup 1 and Setup 2 with an equal number of documents and clusters on the interfaces. Setup 1 (baseline Setup) and Setup 2 (baseline Setup) used the same interface components and features as Setup 3 (experimental Setup), but differed in using either keywords or NE respectively instead of a combination approach (BOW+NE). Therefore, the user had the option to choose which Setup to use and the option to highlight the named entities on the interface using Setup 2 and Setup 3.
We used a dataset that consisted of a collection of documents chosen from TDT2 and TDT3 corpuses which contained 1,468 CNN news documents. We used a copy of the dataset from the
iEvent experiment (Mohd et al., 2012) which involved 28 topics from the TDT2 corpus and 25 topics from the TDT3 corpus. Generally, the number of documents chosen will be adequate for the experiment where the collection’s size is enough to produce clusters to be displayed on the user interface. CNN news stories were chosen because the stories from this source are mostly short, and modelling such stories is, we believe, a good first step before dealing with more complex datasets. Moreover, the dataset used was from 1998 as mentioned above, because users would have no familiarity with or interest in the topics during the evaluation of iEvent.
The users were a mixture of journalists and postgraduate journalism students at the Universiti Kebangsaan Malaysia. Ten users were chosen, of whom three were female. 50% of the respondents were journalists and the remaining proportion were journalism students.
The evaluation of text classification effectiveness in an IR experimental system is often based on statistical measurements. These measurements are precision, recall, and F1-measure (Rijsbergen, 1979; Kowalski, 1997). In our case, the
retrieveddocuments indicated those that had been classified by the system as positive instances of an event and the relevantdocuments were those that had been manually judged to be related to an event.
We treated each cluster as if it were the desired set of documents for a topic. The selection of topics for the Tracking in this experiment were based on the F1-measure (Mohd et al., 2012; Lewis & Gale, 1994; Pons-Porrata et al., 2004). To validate whether the
iEvent interface with combined BOW+NE helps the user to perform the TDT tasks, users were given a bad cluster to track or a bad topic to detect, based on the F1-measure. The chosen topics and clusters given in the user experiment had a combination of good and poor (94.7% - 50.9%) clustering performance according to F1-measure (Mohd et al., 2012).
There were two tasks, Tracking and Detection, which were given to the users.
4.4.1. Tracking Task
The definition of the Tracking task was to track the cluster that contained the identified topic. The user had to track the cluster that included the given topic and show that a sufficient amount of information on the event was provided by the system. This is consistent with the mission of a news journalist. Reporting and Profiling are sub-activities of this task where Reporting is defined as writing an article about a given topic; it requires the user to write an article about a topic through the formulation of significant facts. Meanwhile, Profiling is defined as drafting significant keywords as an outline for a topic. These tasks were considered to simulate natural news monitoring behavior by users. We aimed to create a type of interaction between the users and
iEvent where the latter could perform their own daily news tasks, such as everyday news monitoring and reporting tasks. To facilitate this, we placed the tasks within simulated cases (Borlund, 2002; Borlund & Ingwersen, 1997) shown in Fig. 5, where an example is given of a simulated case in a Topic Tracking task, since the given task’s scenarios should reflect and support a real news monitoring situation.
The steps in carrying out the Tracking task were: A topic summary was given to the users. Based on the information contained in the topic summary, the users were asked to track the related cluster. The users investigated the documents in the related cluster they found. The users were asked to draft the important facts or points to execute the Reporting task. They listed out the dealing cluster. The users created a profile by drafting useful keywords to perform the profiling task. The users expressed their opinion on the features provided by iEvent when performing the Tracking task by completing a questionnaire.
4.4.2. Detection Task
The detection task was the second task performed by the users. It is defined as identifying the topic dealt by a definite cluster. This is consistent with a journalist’s task of identifying significant events that happened on a definite day (Mohd et al., 2012). The steps to carry out the Detection task are:
A specific cluster was given to the users with a list of twenty topics. They were then asked to detect the topics from the documents contained in a specific cluster using any features of iEvent to perform this task. The users were asked to rank a maximum of three topics from the list of twenty topics given, if they felt that the specific cluster contained more than one. Finally, they expressed their opinion on the features provided by iEvent whilst performing the Detection task by completing a questionnaire.
In the Tracking task, there were six topics used and the users were given 15 minutes to complete each topic. They were given a total of 1:30 hours to attempt the entire Tracking task in three sessions.
In the Detection task there were also six clusters given where the users spent a maximum of 10 minutes to complete each cluster. In total, they were given one hour to complete the task in three sessions. Finally, at the conclusion of the tasks, the users completed a questionnaire which took approximately 5 minutes. The whole user experiment took about 2 hours 30 minutes to 3 hours excluding a short training session.
Users were given an opportunity to carry out the tasks using the interface. A Latin Square design (Spärck-Jones, 1981; Spärck-Jones & Willett, 1997; Doyle, 1975) was used which allowed us to evaluate the same topic using different setups. Fig. 6 shows the experimental design, where the order of topics assigned in the Tracking tasks and the order of clusters given in the Detection task were rotated to avoid any learning factor.
Topic 1 (Oprah Lawsuit), for example, has an opportunity to change sequentially from first until sixth in order during the Tracking task. The clusters assigned in the Detection task were hidden during execution of the Tracking task to avoid any intersection of clusters that may have affected users’ performance. The clusters were completed using single pass clustering with the threshold
t=1.48 which created 57 clusters.
During the experiment, 120 tasks were performed. 50% of these tasks were Tracking while the remaining tasks were Detection. The general results showed that 90% of the users liked to use
iEvent in both tasks and 10% of users disliked iEvent because they were accustomed to other news tools. Those who disliked iEvent were all journalists that had an average age of 30-40 years and average working experience of more than 10 years. From the interview session, these participants had previously used news network tools such as Bernama.com.
The results also revealed that more than 70% of the users that used
iEvent in the Tracking task found that the combination approach (BOW+NE) in Setup 3 provided more assistance in conducting the task compared with approximately 17% who found that the keywords in Setup 1 provided more assistance and 13% who found that named entities in Setup 2 provided more assistance. Meanwhile, 90% of all users considered that Setup 3 facilitated the Detection task. During the post-study interviews the users confirmed that the use of a combination of BOW and NE would help journalists to conduct their tasks since it provides significant information on the Who, Where, When, and What of an event at the same time.
A possible explanation for these results might be the users’ success in performing both tasks (see Fig. 7). It is clear that there was a change in the users’ topic interest after using
iEvent across setups. The Kruskal-Wallis Test confirmed that there was no statistically significant difference ( p=0.742) in topic interest before using iEvent across setups. Meanwhile, there was a statistically significant difference ( p=0.010) in topic interest after using iEvent across setups. The users were more interested in a topic in the Tracking task after using Setup 3 ( mean=4.05 sd=0.686) as shown in Fig. 7.
iEvent post-evaluation, the users were asked to circle the setup that they felt was useful in performing the Tracking, Reporting, Profiling, and Detection tasks. The statistics revealed that the use of the combination approach helped to facilitate the user in all tasks as shown in Fig. 8. In the interview, the users told us that the new approach provided them with high quality forms of information where it was more descriptive. This contradicts findings by Mohd et al. (2012) which reported that the BOW approach provided better explanation in the Detection task and the NE approach in the Tracking task. However, in our findings users preferred both approaches to be offered in the Tracking and Detection task since they complement each other. Users receive specific information and at the same time are not missing any relevant information regarding a topic. Thus Setup 3 was significantly helpful in both tasks (Tracking and Detection) because it provided users with broad information (combinations from BOW and NE).
First, the users’ overall opinions on the use of the combination approach (BOW+NE) in Setup 3 were examined. Next, we investigated the users’ performance in the Reporting task, such as the amount of news written, and in the Profiling task, such as the amount and the types of keywords given to write a profile of a story (terms, NE, or a combination). Finally, we investigated whether the users agreed that the use of the combination approach in
iEvent was perceived as useful, effective, helpful, and interesting in performing the TDT tasks.
5.2.1. Overall Opinions
Users’ opinions of iEvent during the Tracking task were analysed between setups. We investigated whether they perceived
iEvent as easy, relaxing, simple, satisfying, and interesting across setups. The Kruskal-Wallis Test confirmed that there was a statistically significant difference in users’ opinion of easy ( p=0.027) in conjunction with the setups. However, there was no statistically significant difference in simple ( p=0.111), satisfying ( p=0.390), relaxing ( p=0.314), and interesting ( p=0.155). Setup 3 was significantly stronger in easy (scale 4, 5) than Setup 1 and Setup 2 during the Tracking task, as appears in Table 3. Scale 1, 2, and 3 are treated as negative due to the zero value that appeared in scale 1 (Linacre, 2006). Meanwhile, scale 4 and 5 are considered as positive. Easy
A ratio of 9 users to 1 found that Setup 3 was easier than the other Setups. 65% of users agreed that Setup 3 (
mean=4.15 sd=0.587) was easy (scale 4) and 25% of them gave scale 5. Interestingly, none of the users found it difficult. This indicates that the use of the combination approach made the Tracking task easier. Relaxing, Simple, Satisfying, and Interesting
As mentioned above, the Kruskal-Wallis Test confirmed that there was no statistically significant difference (p＞0.05) in users’ opinions with regard to relaxing, satisfying, and interesting. It clearly appears, however, that Setup 3 scored higher in terms of scale and ratio compared with Setup 1 and Setup 2, as shown in Table 3.
In the interview session, the users agreed that the use of the combination approach (BOW+NE) provided them with good findings and quick and precise information on the Who, Where, When, and What of an event, which made the Tracking task easy using Setup 3.
Interestingly, the satisfaction factor was also related to the high percentage of correct clusters tracked. This provides strong evidence that
iEvent helped significantly to facilitate users in tracking the correct cluster ( mean=3.98 sd=0.129). The Kruskal-Wallis Test also confirmed that there was no statistically significant difference in the number of correct clusters to be tracked ( p=0.368) in conjunction with the setups. This indicates that the users managed to track the correct clusters using all setups. We classified the correctness of cluster as tracked into four categories:
a. None - where users did not provide any information or they did not complete the task. b. Wrong - where users tracked the wrong cluster. c. Partially Correct - where users list out the minor cluster as their main finding.d. Correct - where users list out the major cluster as their main finding.
The complete Tracking task was successful with 98.33% of tasks as correct and 1.67% as partially correct because sometimes a wrong cluster was chosen that caused wasted time in performance of the task (15 minutes). Moreover, users were confused by some terms, e.g. Merge, Texas, in topics which were highlighted in a cluster on a wrong topic.
The average time taken to conduct this task successfully across setups was approximately 12 minutes, 13 seconds. The users spent about 11 minutes, 35 seconds using the combination approach (BOW+NE) in Setup 3, 11 minutes, 20 seconds using NE in Setup 2, and 13 minutes, 45 seconds for BOW in Setup 1. The time taken in Setup 3 represents a mid-point between the shorter time that was taken in Setup 2 and the longer time in Setup 1. A possible explanation for the mid-time taken in Setup 3 was that it represented a compromise between the specific information (NE) provided by Setup 2 and the general information provided by Setup 1 (BOW). Since Setup 3 provided the users with a meaningful information, both tasks are facilitated.
5.2.2. Reporting Task
The Reporting task is one of the sub activities in Tracking. In this section, we report the findings of users’ performance during this task after analysis of the number of lines that users wrote across setups. There was no statistically significant difference in the amount of news written in conjunction with the setups (Kruskal-Wallis Test,
p=0.536), as shown in Fig. 9. The users managed to write almost the same number of lines in the report on average using Setup 1 ( mean= 7.50 sd=2.819), Setup 2 ( mean=6.45 sd=2.605) and Setup 3 ( mean=7.45 sd=2.762).
5.2.3. Profiling Task
The users were asked to write important keywords as a profile for a topic where there were three types of keywords: named entities, terms, and combination (terms and named entities). For example, users might write keywords such as Exxon (NE), merge two companies (terms), and Exxon merge (combination) for the topic Mobil-Exxon Merger. We analysed the types of keywords provided by the users. The Kruskal-Wallis Test confirmed that there was no statistically significant difference between the types of keywords and the setups used in the Profiling task (
Findings revealed that users had a tendency to provide more terms when they were using Setup 1 and larger named entities when using Setup 2. When they were provided with a combination of BOW and NE in Setup 3, however, the results indicated that they prefer to use NE as the important terms in the profile of the task. Fig. 10 shows that Setup 3 provided users with a larger amount of information (34.74%) compared with Setup 1 (32.83%) and Setup 2 (31.93%). The users interviewed agreed that the new approach (BOW+NE) provided a high quality source of information (balance distribution of terms and named entities), thus it was suitable for use in the Profiling task.
We can conclude, therefore, that the combination of BOW and NE is a good approach in this task. Users are provided with broad terms, and at the same time the combination automatically provides them with highly significant terms (NE).
This section deals with analysing each feature of
iEvent across setups that users perceived as being useful, effective, helpful, and interesting during the Tracking task. Useful
There was a statistically significant difference shown in the evaluation of the ‘CV: cluster labelling’ feature across setups (Kruskal-Wallis Test,
p=0.030). The ‘CV: cluster la-belling’ feature in Setup 3 of iEvent was more useful ( mean=4.15 sd=0.875) compared to Setup 1 ( mean=3.85 sd=1.137) and Setup 2 ( mean=3.30 sd=1.031). It is clear from Table 4 that users found the ‘CV: cluster labelling’ feature useful when they were using Setup 3.
A ratio of 9 users to 1 found that the combination approach (BOW + NE) that was used in the ‘CV: cluster labelling’ feature in Setup 3 of
iEvent was significantly useful with 35% and 55% of users agreeing that it was useful (scale 4, 5), respectively as it seems to show in Table 4.
We can conclude that users found the combination approach (BOW+NE) used in the ‘CV: cluster labelling’ features in Setup 3 of
iEvent more useful than other approaches used in Setup 1 and Setup 2. Effective
There was also a statistically significant difference shown in the evaluation of the ‘CV: keyword approach’ feature across setups (Kruskal-Wallis Test,
p=0.049). Setup 2 was more effective ( mean=4.20 sd=0.523) compared to Setup 3 ( mean=4.50 sd=0.668) and Setup 1 ( mean=3.80 sd=1.105). Users found this feature in Setup 2 and 3 of iEvent more effective as shown in Fig. 11. Helpful
For this opinion value, the Kruskal-Wallis Test also confirmed that there was a statistically significant difference across setups shown in the evaluation of the ‘CV: cluster labelling’ feature (
p=0.011, with mean= 4.30 sd=0.923), in that Setup 3 of iEvent was more helpful than Setup 1 ( mean=3.75 sd=1.251) and Setup 2 (mean=3.25 sd=1.020). Users found this feature was helpful when they were using Setup 3, as shown in Table 5. A ratio of 16 users to 1 found that the combination approach (BOW+NE) used in the ‘CV: cluster labelling’ feature in Setup 3 of iEvent was significantly helpful, with 55% of users agreeing that it was very helpful (scale 5).
This indicates that users found the combination approach that was used in the ‘CV: cluster labelling’ features in Setup 3 of
iEvent more helpful than Setup 1 and Setup 2. Interesting
For this opinion, the Kruscal-Wallis Test confirmed there was a statistically significant difference (
p＜0.05) on two features across setups. These were the ‘CV: top terms’ feature ( p=0.002) and ‘CV: cluster labelling’ feature ( p=0.022).
The evaluation shows that all features of
iEvent were perceived as being interesting in Setup 3 (scale 4, 5) compared to Setup 1 and Setup 2, as shown in Fig. 12. These are ‘CV: top terms’ ( mean=4 sd=0973) and ‘CV: cluster labelling’ ( mean=4.25 sd=0.851) features.
This section is concerned with the analysis of users’ agreement on the ease of detecting the topics and frequency of features with the combination approach (BOW+NE) across setups.
The whole Detection task was conducted successfully with 83.33% of task findings being correct and 13.33% being partially correct. There were 3.33% unsuccessful Detection tasks; we believe the reason for this is because of some users’ confusion between the topics. Generally, this confirmed that
iEvent managed to facilitate user performance in the Detection task, as shown in Fig. 13.
The Kruskal-Wallis Test also confirmed that there was no statistically significant difference in the number of correct topics detected (
p=0.668) across the setups. This proves that the users managed to detect the correct topics using all setups. We classify the correctness of topics detected into four categories:
a. None - where users did not gather any information or they did not complete the task. b. Wrong - where users detected the wrong topic. c. Partially Correct - where users listed out the minor topic as their main findingd. Correct - where users listed out the major topic as their main finding
It was apparent from the Kruskal-Wallis Test that there was no statistically significant difference between the users’ opinions (
p=0.369) on the ease of detecting a topic using setups. However, it seems from Fig. 14 that it was easier to detect a topic using Setup 3 than with Setup 1 and Setup 2.
As mentioned previously in Section 5.1 (General Findings), during the post-evaluation interview the users found that the combination approach (BOW+NE) was more descriptive because they had to detect the topic dealt by a specific cluster in the Detection task, and Setup 3 (combination approach) provided users with Who, Where, When, and What information at the same time.
As shown in Table 6, 55% of users agreed that it was easy (scale 4) to detect a topic using Setup 3 (
mean=4.40, sd=0.699). Surprisingly, none of the users found that it was hard to detect a topic using any of the setups. 85% of users found that Setup 3 makes the Detection task easier than Setup 1 and Setup 2.
Further analysis on the interaction logs proved that users had a higher activity for the ‘DV: document content’ feature in Setup 3 (77.3%) compared to Setup 1 (68.8%) and Setup 2 (65.7%). These results clarified that users clicked less on the ‘DV: document content’ feature with NE (Setup 2) and BOW (Setup 1), than with the combination approach (BOW+NE) in Setup 3. Highlighting BOW and NE in the ‘DV: document content’ feature has provided users with quick and significant information rather than displaying plain document content. In addition, it provided users with links to highlight other clusters which had the NE or BOW in the document content by clicking on the terms or named entities. Additionally, in the successful attempts at the Detection task, the users took less time using the combination approach (BOW+NE) (00:04:22) compared to using NE (00:04:41) and keywords (00:04:50).
Furthermore, the interaction logs showed that there was slight activity in the ‘TV: keyword approach’ feature. This was seen clearly in Setup 3 and less clearly in the other two Setups. This may be due to users attempting to use Who, Where, When, and What terms.
These results indicate that the combination approach (BOW+NE) used in the ‘DV: document content’ feature provided them with all the terms, and the NE used in the ‘DV: document content’ feature helps them to make a quick decision in respect to a topic in the Detection task.
In general, most of the features in Setup 2 (NE) facilitated users in the Tracking task. Meanwhile, Setup 1 (BOW) facilitated them in the Detection task. Interestingly, users agreed that most of the features in the combination approach (BOW+NE) facilitated performance of both tasks. Users found, however, that the use of the combination approach (BOW+NE), for example in the ‘DV: document content’ feature, also provided them with broader information that helped them to detect the topic quickly. This proves that the combination approach (BOW+NE) was effective and efficient in both tasks. Table 7 shows the features with their approaches which facilitated users in performing the TDT tasks.
The Kruskal-Wallis Test confirmed that there was no statistically significant difference between the successful Tracking (
p=0.368) and Detection task ( p=0.668) across setups. It is seems clear that there is consensus that the combination approach (BOW+NE) helped the users in both tasks; moreover, most of them agreed that it was useful and interesting.
Future analysis on the interaction logs enabled us to make a comparison of the mean number of clicks among the successful tasks between setups, as shown in Table 8.
It seems from Table 8 that there is a disparity in mean number of clicks between the approaches that were used in performing tasks across setups. This can be seen clearly between Setup 1 and Setup 2 where mean clicks in Setup 2 in the Tracking task are higher than for Setup 1. Meanwhile, the mean number of clicks in Setup 1 is higher in the Detection task. Surprisingly, it is clear that the mean clicks in Setup 3 was closely equal to the mean average of each task in both Setup 1 and Setup 2, showing that the combination approach (BOW+NE) facilitated both tasks. There are two explanations for the previous findings in interaction logs: First, the higher number of clicks indicates a lack of information or higher interaction between the users and the system; second, the average number in clicks provided by Setup 3 may indicate that users are provided with enough information to perform both tasks.
Users are provided with the two types of information and they can choose to display either keywords or keywords with named entities highlighted. A combination of this information is interesting since named entities are high quality pieces of information and keywords are descriptive. Rather than providing users with the Who, Where, and When it would be interesting to provide the What in one setting and investigate its effects. Thus, this study has presented an experimental study conducted with journalists to investigate the combination of bag of words and named entities (BOW+NE) approaches implemented in the
iTDT interface called the Interactive Event Tracking ( iEvent) system; this includes what TDT tasks these approaches facilitate. The combination approach (BOW+NE) in Setup 3 has provided the users with focused information to facilitate both tasks and enhanced interaction between users and the system. Users provided a large number of terms (BOW and NE) during the Profiling task using Setup 3, which indicates that the iTDT with the combination approach (BOW+NE) is useful in helping journalists to perform their tasks. It has provided users with descriptive and significant information. Overall these findings revealed that the use of a combination approach (BOW+NE) effectively enhanced the iTDT interface and created more interaction between users and the system in both tasks (Tracking and Detection). Journalists require both descriptive information in the Detection task and significant information in the Tracking task. Therefore the combination approach (BOW+NE) in Setup 3 has provided users with focused information to facilitate both tasks and enhanced interaction between users and the system. As a further contribution, this article confirms the value of the combination approach in iTDT which has facilitated the journalists in performing the Tracking and Detection tasks.
[Table 1.] The Differences of the Three Setups
[Fig. 1.] Cluster labelling of the three approaches
[Fig. 2.] Top ten terms of the three approaches
[Fig. 3.] Document content of the three approaches
[Table 2.] Types of Terms in Setup 3
[Fig. 4.] Bag of Words and Named Entities setup (BOW+NE)
[Fig. 5.] Simulated situation
[Fig. 6.] Experimental Design
[Fig. 7.] Box plot of users’ topic interest (scale 1-5) after using iEvent across setups in the Tracking Task
[Fig. 8.] Percentage of users’ preferred setup for the given tasks
[Table 3.] Summary Percentage of Users’ Opinions (easy, relaxing, simple, satisfying, and interesting) Across Setup in Tracking Task
[Fig. 9] Amount of lines written
[Fig. 10.] Percentage for types of keywords across setups
[Table 4.] ‘CV: cluster labelling’ Feature Across Setups Perceived as Useful
[Fig. 11.] Box plot on effectiveness (scale 1-5) of the ‘CV: keyword approach’ feature across Setups
[Table 5.] ‘CV: cluster labelling’ Features Across Setups Perceived as Helpful
[Fig. 12.] Box plot on Interestingness (scale 1-5) of ‘CV: cluster labelling’, ‘CV: top terms’ features across Setups
[Fig. 13.] Percentage of successful Detection tasks
[Fig. 14.] Box plot on Easy to Detect (scale 1-5) with the Setup for the Detection task
[Table 6.] Percentage of User Opinions on Ease of Detecting a Topic Across Setups
[Table 7.] Comparison of Features, with Approaches, Which Facilitated TDT tasks
[Table 8.] Comparison of Features, with Approaches, Which Facilitated TDT tasks