PALTask chat tool


Personalized web-tasking seeks to improve the user’s experience through the automation of repetitive and ordinary tasks to fulfill personal goals. An example of an ordinary task is the search and selection of proper web-resources based on a specific matter of concern. However, in a dynamic situation, such as an online conversation between two or more users, the topics are dynamic and uncertain. This implies for the users to constantly switch
among web sites, formats and keywords, in order to search relevant information while simultaneously carrying out the conversation.

Automating the user’s personalized web tasks in this scenario is a challenge due to the lack of understanding of the user’s dynamic context. Moreover, the user’s feedback during the execution of the task is required to refine the search and the resources retrievalTo achieve automation of the web-resources listing task, our proposed tool relies heavily on context analysis. This context is obtained from several sources: 1) the users’ personal context spheres, and 2) the chatting contents itself.

  • Andi Bergen
  • Pratik Jain
  • Lorena Castañeda
  • Hausi Müller

Technical details of PALTask chat

As depicted in Figure 1, PALTask is composed by two main elements. In the left a traditional chat interface, and in the right the resources list display. At the same time, this resources display provide user's interactivity through some operations, such as format filtering, approve or disapprove a resource, and sharing.

Figure 1. PALTask User's interface proposal

Figure 2 shows the interactions of PALTask numbered 1 to 5. 
  1. The user interacts with PALTask in the form of an online conversation---or chat. 
  2. The text from the chat is extracted and then analyzed by the keyword extractor to get the topics of the conversation.
  3. The user's context sphere (all her preferences, previous web interactions, and any contextual information related with her presence in the web) is used to assess the topics of the conversation and filter those highly related with the user's interests and personal goals.
  4. the final keywords from the conversation and the user's personal context are queried in the web to retrieved the proper resources. The extensible module represents the cloud multiple resources available, for example videos, documents, and music.
  5. Finally, the results are parsed and displayed in the PALTask user's interface.

Figure 2.  High level interactions (click to enlarge)

Figure 3, describes the component architecture of PALTask chat tool. The chat tool elements (GUI, Client, and Server) are a basic implementation in Java using traditional communication protocols and techniques. The Keyword extractor is a python based implementation of RAKE. Finally, the Personal Context Sphere is a XML representation of the proposal of Villegas et al.

Figure 3. PALTask component architecture (click to enlarge)

How does it look? Here are some screenshots

Screenshot 1. Both users login (click to enlarge)

Screenshot 2. First conversation topic (click to enlarge)

Screenshot 3. Both users conversations topics (click to enlarge)

Screenshot 4. sharing resources from one user to the other (click to enlarge)