Penn State University - TexPlorer
VAST 2007 Contest Submission

Authors and Affiliations:

Chi-Chun Pan, Penn State University, cpan@ist.psu.edu

Anuj R. Jaiswal, Penn State University, arj135@psu.edu

Junyan Luo, Penn State University, jluo@psu.edu

Anthony Robinson, Penn State University, acr181@psu.edu

Student team: [*] YES [  ] NO
If you answered yes, name the faculty who agreed to be your sponsor: Name, email address

Prasenjit Mitra, Penn State University, pmitra@ist.psu.edu

Alan M. MacEachren, Penn State University, maceachren@psu.edu

Ian Turton, Penn State University, ijt1@psu.edu

Tool(s):

TexPlorer uses the following tools to process and visualize the VAST 2007 contest dataset:

1. FactXtractor is a named entity and entity relationship extractor developed by the North-East Visualization and Analytics Center at the Pennsylvania State University. FactXtractor processes text documents using GATE and indentifies entity relations with both syntactical and semantic analysis.
2. ConceptVISTA is an ontology creation and visualization tool developed by researchers at the GeoVISTA Center at the Pennsylvania University. We use ConceptVISTA to visualize concept maps extracted by FactXtractor. More information about ConceptVISTA can be found at http://www.geovista.psu.edu/ConceptVISTA/.
3. MEAD is a public domain portable multi-document summarization system original developed at the University of Michigan. We use MEAD to create summary for text documents and document clusters. More information about MEAD can be found http://tangra.si.umich.edu/clair/mead/.
4. CLUTO is a family of computationally efficient and high-quality data clustering and cluster analysis programs developed by the Digital Technology Center (DTC) at the University of Minnesota. We use CLUTO to compute content-based document clustering. More information about CLUTO can be found at http://glaros.dtc.umn.edu/gkhome/views/cluto.
5. SIMILE Timeline is a DHTML-based AJAXy widget for visualizing time-based events developed as part of the SIMILE project at MIT. More information about the SIMILE Timeline can be found at http://simile.mit.edu/timeline/.
6. WordNET is large lexical database of English developed at Princeton University. We use WordNET to perform semantic expansions of keywords within our document filtering tools. More information on WordNET can be found at http://wordnet.princeton.edu/.

Data set used:   [ * ] RAW DATA SET     [   ] PRE-PROCESSED  SET

TOC:  Who – What – Where – Debriefing - Process - Video – Video (High Resolution)

(ADD your links to the video – use a relative link so that it works everywhere)

---

1. WHO: who are the players engaging in questionable activities in the plot(s)?   When appropriate, specify the association they are associated with

Name	Associated organization	Involved in illegal activities? (Yes/No)	Involved in terrorist activities? (Yes/No)	Most relevant source files (5 MAX)
Faron Gardner	Animal Justice League	Yes	Yes	Week-of-Mon-20030818.txt_23, Week-of-Mon-20030602-1.txt_66, Week-of-Mon-20030714-2.txt_25 chinchilladreamin.blogger.com
Cesar Gil	SHAC	Yes	No	Week-of-Mon-20040705.txt_86, Week-of-Mon-20030609.txt_4, Week-of-Mon-20030901-1.txt_36, chinchilladreamin.blogger.com
Catherine Carnes	SPOMA	No	No	Week-of-Mon-20030609.txt_4, Week-of-Mon-20030818.txt_23, Week-of-Mon-20030526-2.txt_57
Daniel Andreas San Diego	Revolutionary Cells, SHAC	No	Yes	Week-of-Mon-20031006-4.txt_26, Week-of-Mon-20031027.txt_7, Week-of-Mon-20031006-5.txt_28, Week-of-Mon-20031202-3-5.txt_82, Week-of-Mon-20040223-2.txt_25
Kevin Jonas	SHAC	Yes	No	Week-of-Mon-20040223-2.txt_25, Week-of-Mon-20040531-1.txt_33
John Burton Wade, Aaron Linus, Adam Blackwell	ELF	Yes	Yes	Week-of-Mon-20040412-3.txt_38
William Cotrell	ELF	Yes	Yes	Week-of-Mon-20040322-1.txt_43, Week-of-Mon-20040329-4.txt_85, Week-of-Mon-20040614.txt_27, Week-of-Mon-20040426-3.txt_34

---

2. WHEN /WHAT:   What events occurred during this time frame that are most relevant to the plot(s)? 

Provide a text list of events following the sample layout.  Use short description (i.e. one or 2 lines per event)

Provide what you think is the best subset of events (20 events MAX)

	Date Can be a range	Event description	Most relevance source files (5 Max)
1	Jun 2003	AJL smashes cash registers, broke windows at Pet Smart	Week-of-Mon-20030602-1.txt_66, Week-of-Mon-20030609.txt_4
2	Jul 2003	AJL leaves package at LA times office, claims meat has been poisoned at 20 supermarkets in LA.	Week-of-Mon-20030714-2.txt_25
3	Sep 2003 – Jun 2004	Possible advertisement shows Cesar Gil is in the US and selling chinchillas including the short-tailed variety. His blog shows he is still active in chinchilla breeding in Jan.	Week-of-Mon-20030901-1.txt_36, chinchilladreamin.blogger.com
4	2003-2004	Stop Huntingdon Animal Cruelty (SHAC) engages in acts of intimidation and threats against companies working with Huntington Life Sciences.	Week-of-Mon-20040223-2.txt_25, Week-of-Mon-20040202-2.txt_70, Week-of-Mon-20040607-3.txt_25, Week-of-Mon-20040524-1.txt_67,
5	August 2003 – Nov. 2003	Attempted bombing at a west Michigan water bottling plant, ELF claims credit.	Week-of-Mon-20031110-1.txt_54,
6	August 2003- Sept 2003	Bombing at Chiron, Emeryville and Shaklee, Pleasanton, Revolutionary cells claims credit.	Week-of-Mon-20031006-4, Week-of-Mon-20031027.txt_7, Week-of-Mon-20030825-4.txt_30,
7	Sept 2003	Vandalism of a Chiron executive's car and the trashing of a biology lab at Louisiana State University	Week-of-Mon-20031027,
8	October 2003	E-mails after the bombings show reason of attack is association with Huntingdon Life Sciences (HLS). HLS also targeted by SHAC.	Week-of-Mon-20040621-1.txt_29, Week-of-Mon-20031006-5.txt_28
9	October 2003 – February 2004	FBI searching for bombing suspect Daniel Andreas. Alleges can be responsible for more attacks.	Week-of-Mon-20031006-4.txt_26, Week-of-Mon-20031027.txt_7, Week-of-Mon-20031006-5.txt_28, Week-of-Mon-20031202-3-5.txt_82, Week-of-Mon-20040223-2.txt_25
10	Feb 2004	Chiron, FBI alleges SHAC, ELF, ALF hand in Chiron Bombing, alleges link between Kevin Jonas and Daniel Andreas.	Week-of-Mon-20040223-2.txt_25, Week-of-Mon-20040614-1.txt_2
11	Feb 2004	Oxford University Animal Research Facilities declared legitimate targets by ALF	Week-of-Mon-20040209-2.txt_97
12	May 2004	The federal indictment unveiled yesterday against Stop Huntingdon Animal Cruelty	Week-of-Mon-20040524-1.txt_30
13	May 2004	members of Stop Huntingdon Animal Cruelty, indicted to conspire to commit animal enterprise terrorism	Week-of-Mon-20040614-1.txt_3
14	Jun 2004	Shareholders in Montpellier, the construction group building new research laboratory for Oxford University, received threatening letters	Week-of-Mon-20040621-1.txt_29, Week-of-Mon-20040621-4.txt_48
15	Jun 2004	ALF, ELF declares targeting building developers, car dealerships legitimate.	Week-of-Mon-20040607-3.txt_24, Week-of-Mon-20031013-6.txt_58
16	Sep 2003 - Jun 2004	Earth Liberation Front claims responsibility for arson at housing developments, car dealerships.	Week-of-Mon-20040607-3.txt_70, Week-of-Mon-20040607-3.txt_24, Week-of-Mon-20040126-5.txt_88, Week-of-Mon-20040614.txt_27, Week-of-Mon-20040216-3.txt_62
17	Jun 2004	Rapper r’Bear receives animals for his animal ranch including short-tailed chinchillas.	Week-of-Mon-20040614.txt_94
18	Jul 2004	Post on Cesar Gil blog relating to monkey pox outbreak. Posts identifying with Faron Gardner and Collie Carnes.	chinchilladreamin.blogger.com
19	Jul 2004	Second Monkey pox outbreak in LA area. Two people die. Chinchillas identified as source. Cesar Gil sought in Monkey Pox outbreak. Believed to have fled country.	Week-of-Mon-20040705.txt_83, Week-of-Mon-20040705.txt_86
20 max	Jul 2004	Rapper r’Bear falls seriously ill and is rushed to hospital.	Week-of-Mon-20040628.txt_61

---

3. WHERE: What locations are most relevant to the plot(s)?

Follow this example layout.  Use only one-line per item.

	Location	Description	Most relevance source files (5 Max)
1	Pleasanton, CA	Bombing	Week-of-Mon-20031006-4.txt_26, Week-of-Mon-20031027.txt_7, Week-of-Mon-20031006-5.txt_28, Week-of-Mon-20031006-5.txt_28,
2	Emeryville, CA	Bombing	Week-of-Mon-20031006-4.txt_26, Week-of-Mon-20031027.txt_7, Week-of-Mon-20031006-5.txt_28, Week-of-Mon-20031006-5.txt_28,
3	Los Angeles, CA	Monkey pox outbreak, AJL Attacks.	Week-of-Mon-20040705.txt_83, Week-of-Mon-20040705.txt_86, Week-of-Mon-20030602-1.txt_66, Week-of-Mon-20030609.txt_4, Week-of-Mon-20030714-2.txt_25
4	San Francisco, CA	Recent Bombings.	Week-of-Mon-20031006-4.txt_26, Week-of-Mon-20031027.txt_7, Week-of-Mon-20031006-5.txt_28, Week-of-Mon-20031006-5.txt_28,
5 max

---

4. DEBRIEFING

Radical Attacks on Huntington Life Sciences

Huntington Life Sciences Targeted

Huntington Life Sciences is systematically targeted by radical animal rights groups such as SHAC, ELF, ALF and Revolutionary Cells. Even organizations having business ties with HLS are acceptable targets. SHAC has targeted HLS and organizations working with HLS by intimidation and threats. Employees of Chiron Corp. are targeted by SHAC. Recent attacks include bombings at Chiron Corp, Emeryville and Shaklee Corp in Pleasanton.

ALF and ELF have been involved in bombing of a Vail Ski Resort and a failed bombing at Detroit. Further arson and vandalism at SUV showrooms has been attributed to these organizations. William Cottrell and Josh Connole are charged in various arson events and alleged to be ELF members.

In addition, Chiron Corp has alleged that the FBI has indicated that SHAC is directly involved with bomber Daniel Andreas and provides information that SHAC president Kevin Jonas is suspected to know the bomber. Further, the company sued SHAC for harassment and intimidation. Recent indictments of SHAC members for harassment and intimidation, four grand jury investigations against SHAC and a federal indictment against SHAC, ALF and ELF are cause of concern about plans being developed for near term attacks.

Bomber on the run

Daniel Andreas San Diego is wanted in regard to the recent attacks on Chiron Corp and Shaklee Corp. An FBI analyst outlines similarity with the Una-Bomber. He is on the run and is considered armed and dangerous by the FBI.

The SHAC association with ALF, ELF, etc

In a recent FBI conference, the deputy director named SHAC, ELF and ALF as special interest terrorist organizations. Further, Chiron Corp brought a lawsuit against SHAC alleging ties between the Daniel Andreas and SHAC president Kevin Jonas. Recent indictments of members of SHAC cast doubts whether the organization is involved in violent attacks despite statements to the contrary. FBI links recent environmental terrorist activities with SHAC, ALF and ELF, thus it is possible that SHAC members are collaborating with members of ALF and/or ELF.

In the crosshairs

ALF declares that Oxford University is a valid target due to links with HLS. This statement is a cause for concern due to the similar pattern of attacks as well as ties of Oxford University with HLS. Recent bombings at Chiron Corp. and Shaklee Corp. were preceded by vandalism and threats. Recent intimidation tactics against shareholders of Montpellier seem to be tactics of SHAC. SHAC USA, however, has been served a federal indictment as well as four grand jury investigations in regard to terrorist activities. Attacks against organizations having ties with HLS have been brutal and this threat must not be taken lightly.

Monkey Pox Outbreak

Cesar Gil

Cesar Gil is a chinchilla breeder and fanatical animal activist as mentioned by himself in his blogs over a period of time between Sep 2003 and Jun 2004. His blog has a post under the title “Chinsurrection” depicting chinchillas and referring to the monkey pox outbreak. In Sept 2003, Cesar Gil is selling chinchillas including the short-tailed variety. In July 2004, a second monkey pox outbreak strikes Los Angeles. Officials at the Center for Infectious Diseases identify chinchillas as the source of infection. Officials are also searching for the Cesar Gil in connection to the monkey pox outbreak. Officials believe that Cesar Gil is currently on the run and fled the United States.

AJL/ SPOMA involvement

AJL has been actively involved in illegal incidents such as the raid on 3 Pet Smart stores and the alleged poisoned package and letter left at the LA Times office, claiming poisoned meat in 20 supermarkets in LA during Jun 2003. In Cesar Gil’s July 2004 posts on his blog, he references Faron Gardner (AJL) and Catherine Collins (SPOMA) as friends in addition to other individuals. In addition, Cesar Gil refers to the activities of AJL and Faron as the only way animals can be defended. Faron Gardner’s previous rants (August 2003) regarding an armed struggle is cause for concern.

Dangerous Signs

Rapper r’Bear is working towards creating an animal hunting ranch. In Jun 2004, he receives shipments of a number of animals including the short-tailed chinchillas. Cesar Gil had posted advertisements for sale of short-tailed chinchillas in Sept 2003. In Jul 2004, rapper r’Bear falls seriously ill and has to be rushed to the hospital.

The Global Ways connection

In October 2003, a news story had a reference to a user falling sick after handling a shipment of catfish which was imported through Global Ways. Global Ways responded by blaming the contamination to an experienced packer in South America. In January 2004, the Fish and Wildlife Services issued a warning with regard to contaminated catfish shipments from South America. Global Ways was identified as a possible source. More investigation of Global Ways records is advised though these events do not seem to be tied to a planned eco-terror event. However, Madhi Kim, CEO Global Ways, and his connection to rapper r’Bear (in Mar 2004) and the recent completion of the lake at r’Bear’s ranch raises more questions regarding contamination at the animal ranch “Shravaana”.

---

5. VISUALS and Description of ANALYTICAL PROCESS

Our first step was using FactXtractor, a name-entity and entity-relationship extractor developed by the North-East Visualization and Analytics Center at the Pennsylvania State University. FactXtractor was, run over all the documents (news stories, blogs etc.). FactXtractor processes text documents using GATE and indentifies entity relations with both syntactical and semantic analysis.

Our next step was document filtering using semantic expansion by use of the WordNET dictionary. We input a set of keywords such as “arson”, “terrorism”, “drugs”, “police”, “indictment”, etc. which encompassed our problem space. These keywords were used to find semantically similar keywords by using the hyponym relationship from the WordNET dictionary. Our keywords were expanded to a total of about 200 generalizations. We then filtered out documents which did not contain these words. Our resulting document space was thus halved.

We then ran CLUTO to compute content-based document clustering. The document filtering described above should improve the quality of our cluster since only relevant documents are processed. Figure 1 shows our content-based clusters for the VAST dataset. The blue hyperlinked numbers show the cluster numbers and the link providing access to the cluster information. Within each cluster, the keywords in bold are the cluster features which represent the set of high frequency keywords which appear most often in documents with that cluster while the numbers next of the features show the frequency of occurrence. For example, for cluster #25, the keyword “fish” appears with high frequency (16.6%) in the set of documents.

Figure 1: Content Based Clustering for the VAST Dataset. (Only first 9 features are shown. Features shown are stemmed keywords. Stemming is based on the Porter Stemming Algorithm. For example, in cluster #27, the keyword “consume” is shown as the stemmed version “consum”. Note: the string “consume”, “consuming” etc are all stemmed to the string “consum”). Cluster #25 is of interest since the keyword “bomb” is present. Similarly, cluster #20 looks interesting in relation to drugs and tropical fish.

Figure 2 shows a snapshot of TexPlorer. The top panel contains a timeline view which shows documents arranged temporally (those in the cluster are highlighted in red). The left bottom panel contains buttons to view the generated content-based cluster features (Feature List Tab), tree view of the cluster hierarchy (TreeView Tab) and the search utility (Search Tab). Currently the Search Tab is selected. The right bottom panel contains the window where detailed information on the entity feature classes can be reviewed.

Figure 2: TexPlorer snapshot showing the timeline view (top panel), search panel (bottom left panel) and cluster information panel (bottom right panel)

Our next step involved finding a relevant cluster since the cluster keywords show only the high frequency keywords that appear in documents. We at this point decided that “bombing” or related keywords would be an interesting direction to explore. Parsing through the cluster keywords we found that the keyword “bomb” appeared in cluster #25. However relevant information might be hidden and using the cluster hierarchy by itself is not sufficient to uncover it. For example, other “bombing” related documents might occur in different clusters and might not appear in that cluster’s keywords. For example other bombing (or bomb) related documents might have different semantically similar keywords such as “grenade”, “dynamite”, “explosives” or “Molotov cocktail”. By using our text search utility, as shown in Figure 2, we searched for relevant terms (i.e. “bombing”) which returned a list of documents as shown in Figure 3.

Figure 3: Searching for keyword “bombing” reinforces the view that cluster #25 is the relevant cluster since 8 out of 11 documents were placed there.

We explored the results returned by our search utility and found that 8 out of 11 documents containing the keyword “bombing” were placed in cluster #25. This reinforces the view that cluster #25 is the cluster best representing bombing events. We then decided to explore this cluster to build an initial hypothesis. On clicking the cluster #25 link, our tool expands the cluster hierarchy and brings cluster#25 into focus. The relevant entity information derived by FactXtractor contained in the cluster (i.e. Persons, Locations and Organizations), are displayed in the right hand bottom panel, the cluster hierarchy in the left hand bottom panel and a timeline display tool on the upper panel. Figure 4, shows a screenshot of our tool with the cluster hierarchy expanded.

Figure 4: TexPlorer screenshot, with cluster #25 in focus. The upper panel shows the temporal view of documents. Documents with a red tab refer to documents in cluster #25; else they are shown with a blue tab. The left hand bottom panel shows the cluster hierarchy. Notice that cluster #25 and #9 have the same parent #39. The filelist tab can be expanded to view all files in this cluster.

On clicking the cluster #25, the entity feature classes, a summary of documents within and the cluster features are loaded into the left hand bottom panel. Figure 5a shows the automatic summarization of text documents within this cluster and the cluster feature keywords. Figure 5b shows the three entities (persons, locations and organizations) (Note: The system by default shows 5 entities per feature class. The user can adjust the list to show up 25 entities; the figure lists only 2 to allow a representative sample to be shown.).The person entities are shown with affiliations if sufficient information was present in text during the FactXtractor name-entity extraction process. With each entity, the top 5 files that it appears in are listed (if it appears in that many).

Figure 5a: Snapshot showing automatically generated summary, feature list and the extracted person entities for cluster #25 (top two entities are shown). For entity tables, the +/- button allows an analyst to override the system’s ranking of results (either promoting an entity to the top of the list or removing it from the list of key entities). The three columns in the People entity tables show the Entity Name, Entity Affiliation (if available in text extraction) and top 5 ranked documents (if more than five documents for an entity exist).

Figure 5b: This snapshot shows the three feature classes extracted from the text documents. The Person entity feature class represents names extracted, the Organization entity feature class represents extracted organizations, and Location entity feature class shows extracted locations. Red entities are entities which have been tagged by a user to be important.

The above step was followed by evaluating the extracted entities by viewing the documents. An analyst can quickly evaluate content by using our text preview by moving the mouse over the file lists (Figure 6a) or by viewing the text in its entirety by clicking on a file link (Figure 6b). Content is shown with all the entities (persons, locations and organizations) highlighted and color coded which allows for quick review of information. For example, by using these tools, we were quickly able to grasp that “Daniel Andreas San Diego” and “Kevin Jonas” were important people who were involved in some illegal/ terrorism activities.

Figure 6a: Text Preview for the Person “Daniel Andreas” showing the extracted entities. “Daniel Andreas” was ranked the first by the system. “Kevin Jonas” was ranked lower but a user found him to be of interest and promoted his score. Similarly, the organizations “Chiron Corp” and “Shaklee Corp.” have been tagged as important.

Figure 6b: Complete Text view for the Person “Daniel Andreas San Diego”. The document is highlighted and color coded with the different entity classes (persons, locations, organizations and time) allowing easy review of information.

When the cluster #25 was clicked, the top panel containing the timeline tool is loaded with the timeline of the documents. Further while reviewing the complete text view, we can move the focus of the timeline to the current document by clicking on the “move Timeline to” button. Figure 7a shows the result of this operation. By using the timeline utility we were quickly able to view the documents in the temporal neighborhood (as shown in Figure 7a, 7b and 7c) where the links between recent terrorist attacks and organizations can be easily reviewed.

Figure 7a: Moving the timeline (red tab in the top panel) to the temporal neighborhood of the document being reviewed (bottom right panel). Documents in the current cluster are shown with a red tab while documents in other clusters are shown with blue tabs. The current document is viewed in the timeline by clicking on the red tab which reveals an automatically generated summary of this document.

Figure 7b: Exploring the timeline by viewing an article from a different cluster allows an analyst to quickly make a link between the recent ELF attacks and bombing events on Chiron Corp and Shaklee Corp.

Figure 7c: Exploring the timeline view further by viewing another document from the same cluster in the temporal of the first document (shown in Figure 7a) allows an analyst to quickly find the relationship between about ELF being declared a terrorist organization and its recent arson attacks.

However, entities can appear in multiple clusters, i.e. the organization entity “Chiron Corp.” can appear in documents which have been clustered elsewhere. To explore these relationships, each entity can be clicked. On clicking this entity, all documents in all clusters matching this entity are returned. On clicking “Chiron Corp.”, we found that this entity occurred in Cluster #9 in addition to cluster #25 as shown in Figure 8. On exploring, cluster #9 we were able to make connections that “Chiron Corp.” has business associations with “Huntingdon Life Sciences”, the “ELF” and “ALF” vandalism attacks, SHAC intimidation and harassment tactics etc.

Figure 8: Entity search revealing links to information in other clusters by clicking on entity “Chiron Corp.” in cluster #25. The document with fileid 155 allowed us make the link that Huntington Life Sciences had association with Chiron Corp. in addition to the intimidation tactics by the group SHAC.

Since entity extraction is not perfect, our tools allow an incorrect entity (for example “Washington” labeled incorrectly as a person) to be quickly corrected by use of drop down menus as shown in Figure 9. Further, on finding an entity of particular interest we can boost its relevance by clicking on the “+” button. This causes the entity to be color coded red as shown in Figure 6 where the entities “Daniel Andreas San Diego”, “Kevin Jonas” are shown highlighted in red. In addition, if an entity is found to be not of any importance in our hypotheses, it can be demoted by using the “-” button as shown in Figure 6. This information is carried over globally to all clusters, allowing fine tuning of the entities of interest and building towards our hypotheses.

Figure 9: Entity Correction for “Washington” incorrectly identified as a Person.

For person and organization entities, we can visualize a concept map by clicking on the “show concept map” buttons under either the Person panel or Organization panel. TexPlorer writes the corresponding OWL (Ontology Web Language) files which can then be viewed within the ConceptVISTA application as shown in figure 10. Concept maps are extremely useful and intuitive visual tools which provide an effective mechanism for visualizing information. The OWL language provides reasoning mechanisms which can prove useful for deducing answers from the information stored.

Figure 10: ConceptVISTA snapshot with the right panel showing the concept/ node of “Daniel Andreas San Diego” expanded to show the information captured by TexPlorer. The left hand bottom panel shows the various entity classes (for e.g. Person, Organizations etc.) with the Person entity class expanded to show the different persons present in the concept map for cluster #25.

By using the concept map, we were able to quickly find that “Daniel Andreas”, “Kevin Jonas”, “FBI” and “Shaklee Corp.” are referred to in the same article as shown in Figure 11a. Similarly from Figure 11a, we can see that Daniel Andreas is involved with the organization “Shaklee Corp.” and is wanted by the “FBI”. Viewing the article of interest within ConceptVISTA lets analysts make more informed decisions on the possible relationships. Figure 11b shows the text article of reference (center of bottom of Figure 11a).

Figure 11a: Concept map of Person and Organization entities viewed in ConceptVISTA showing the link between Daniel Andreas, Kevin Jonas, Shaklee Corp. and the FBI in the article below (center of bottom of picture). We can also see that the FBI is searching for Daniel Andreas and much more.

Figure 11b: Viewing the text in ConceptVISTA allows analysts to view the story and make more refined estimates of the underlying information.

Figure 12a, shows the concept map for the entity “Virgil Butler” and clearly shows he was fired from the organization ”Tyson” and also made some claim in regard to the organization “Tyson”. In addition, the news article of interest is shown in Figure 12b.

Figure 12a: Concept neighborhood of the Person “Virgil Butler” showing the locations, organizations, people and new articles of interest. The links between Tyson, Virgil Butler, and the article of interest are clear.

Figure 12b: Viewing the article of interest within ConceptVISTA allows making the links between different entities much clearer. Notice that the text shows exactly the visual links shown in Figure 11a.

We can then visualize the important locations on a map, by using the “Show World Map” button as shown in Figure 13. The number of location shown on the map can be varied by using the drop down list in the Important Location table.

Figure 13: Map visualization showing the 10 most relevant/ important locations for cluster 25.

TOC:  Who – What – Where – Debriefing - Process - Video – Video (High Resolution)

(ADD your links to the video – use a relative link so that it works everywhere)