Overview / Usage

In the “Internet” era, search engines are widely used as gateways to people and the Internet. But most of the time, users can't build the initial query very well. We are developing a web page to help users build the initial quary.

Methodology / Approach

Based on the traditional search engines, OptimNet has added a function which named “Target Vocabulary Recommendation”. We can guess the intent based on the keywords entered by the users and return to the target vocabulary for selection. Through constant iterative search, the user is gradually guided to improve the description of the search target.

For the first time, the analogical reasoning thought is integrated into the search engine service. The contextual features of the vocabulary implied by the word vector are used in the search engines ,refering the way of people exampling. It provides a new solution for the fuzzy search field.

OptimNet provides users with three search modes based on the amount of information they have. Based on the user's original search habits, OptimNet gradually guides users to improve the input keywords, which alleviates the information asymmetry between users and search engines, so that users can obtain the required content more efficiently and accurately.

Search Aggregation: If users can accurately describe their search targets, they can enter them into the search box for accurate search. OptimNet obtains information from multiple search engines. It filters out duplicate results and sorts them, and then returns results that are more appropriate to the user's expectations.

Analogical reasoning: If users cannot effectively describe the expected vocabulary, the analogy can be used as an example. OptimNet uses the implicit features of objects to perform feature operations. It allows search engines to infer target vocabulary that is closer to the user's expectations.

Target vocabulary recommendation: After users make a vague input, OptimNet guesses the users’ intention and recommends the target vocabulary by analyzing and understanding the users’ input. Users can add a more detailed description of the expected vocabulary by selecting the relevant target vocabulary provided by OptimNet.

Technologies Used

■Word2Vec：The vector of the word is understood as the contextual feature of the word, or the characteristics of the vocabulary itself, so that the vocabulary vector is operated to infer the resulting vocabulary.

■Dependency parsing analysis：Use dependency syntax analysis to understand sentences and effectively improve the accuracy of target words.

■Named entity recognition technology：The user's target result is a named entity such as name, place, organization, etc. This technology can speed up the search and narrow the search range.