Development of an Interactive Product Search System with Aspect-Based Opinion Filtering

Status


Finished bachelor thesis

Formalia


Targeted audience
  • AI Bachelor
Preconditions
  • Ability to read and understand papers written in English.
  • Ability to perform academic writing.
  • Strong backend programming and scripting skills (e.g. Python essential)
  • Web development skills (essential)
  • Lecture Internet-Suchmaschinen or Information Mining (desired)

Task description


Most online shops enable users to comment on offered products. User reviews contain a high value for other users — especially for laymen — as they characterize the pros and cons of products and thus are potentially helpful resources for deciding to buy an item. However, popular products can have hundreds of reviews leading to the problem that manual scanning of these reviews is too exhausting and time consuming for the majority of the users.

The goal of this thesis is to develop an interactive search system for product search which takes into account extracted information from user reviews for the retrieval of products. The application domain is the Laptop domain. At first, meta-data filtering of items, i.e. filtering based on product characteristics, should enable users to match the target topic. Aftwerwards another filtering process is used to tailor the result list based on crowd opinion. Products that do not (partially) satisfy the crowd w.r.t. some aspects are filtered out or marked appropriately in the result list. Another requirement is to visualize the result items transparently to the user. This means one should receive feedback for each entry about the meta-data conditions as well as opinion-based aspects criterions. Futher weighting of the conditions allows to re-order the result list. In terms of information extraction of user reviews, we aim to use State-of-the-Art tools for aspect extraction and sentiment analysis. These models serve as basis for the opinion-based filtering stage.

In a former project Voice Queries in Product Search with Vague Conditions a meta-data filtering search system for Laptop products was developed. This search system is used to extend it with the opinion-based fitlering stage.

This thesis includes the following phases:

  • Literature scan. This should be done before the actual project starts. Here the student will be given some initial papers. Based on these papers, the student should collect more papers, perform a review of all the papers and prepare an oral presentation of 30 mins.
  • Data acquisition of products and its reviews (including crawling and indexing).
  • Usage of State-of-the-Art tools for aspect extraction and sentiment analysis in the Laptop domain.
  • Implementation of an Interactive Search Interface with preference-based manipulation and transparent result visualization.