SHANTI - Sciences, Humanities, and Arts Network of Technological Initiatives

Topical Research Information Modules (TRIM) / Profile

Home Page:

Topical Research Information Modules (TRIM) should fit your needs when you have permanent and/or in-depth research topics. TRIMs can be viewed as specialized semantic modules, tuned up to your topic and preloaded with relevant documents to be analyzed. Unlike the iResearch Reporter, TRIMs are typically used without a search query. Once a documents set has been analyzed, the same TRIM can be loaded with a fresh portion of documents, collected using a modified search query. TRIMs provide multi-perspective, panoramic view of information in your research area.


Two main TRIM varients: 

1. TRIM-summarization


  • Multi-document summarization with semantic search: Semantic search capability is applied to identification of text passages needed to be excerpted from the relevant documents. Semantic search implementation is based on the dynamically created lists of concepts, characteristic for the user-indicated topic and its automatically identified subtopics. E.g., for the topic GUN VIOLENCE the system would apply such concepts as “shootings”, “firearm related deaths”, and for the topic GUN RIGHTS  - such concepts as “Second Amendment”, “keep and bear arms”, “gun restrictions”.
  • Breakdown of the summarized information according to the main theme subtopics: The TRIM way of multi-document summarization includes addressing the topic area as a whole, as well as with regards to its specific perspectives (subtopics). The same documents set is reprocessed in order to obtain specific information coverage from each of multiple perspectives.
  • Organizing the extracted information: The collected passages are ordered by their semantic closeness. The resulting summary is subdivided into two sections basing on the degree of pertinence. These sections – Most Pertinent and Possibly Useful – are structured further, introducing thematic sections and paragraphs, thus making the summary even more well-readable.
  • Assigning headings to thematic sections: The basic requirement for selecting a section heading is that it should represent a specific concept term most frequently occurring in a given text section. Its synonyms and related concepts are taken into consideration as well. Glancing through the created table of contents, the user can quickly glean the overall thematic structure, allowing him/her to quickly access a section of interest.
  • Ranking the source documents by relatedness to topic & subtopics: The documents related to each subtopic or the theme as a whole are ranked according to amount of specific informative passages they contain.
  • Updating the documents set: Once a documents set has been analyzed, the same TRIM can be loaded with a fresh portion of documents, collected using (optionally) a modified search query.


2. TRIM-categorization

  • Categorization: The search results (retrieved documents) are organized into thematic groups (categories). The categories are NOT predefined, but established dynamically – they are labeled by specific concept terms semantically close to the user’s query. 
  • The categories set is created with the concept-based approach: Categories represent the most core and widely used concepts occurring in the publications on the query subject. Such categories are identified through a cascade-like multi-stage mechanism involving search and analysis of information found on the Internet and/or other sources according to a user’s topic in general. The candidate categories are further refined by assessing their significance in documents found according to a user’s query. This way, the retrieved relevant documents are categorized by concepts, which are core for a given theme, as verified on large masses of information.


Categorization  Example:

Categories:CLEAN ENERGY 













  • Ranking the documents: The documents assigned to each category are ranked by their amount of category-specific content. Each document can be assigned to more than one category – typically to many. 
  • Summarization: For every category, a user can get a cumulative summary of all documents fitting into this category, and also obtain the category-specific information for every document independently. 


Interface Languages
English (416)