Bidirectional LSTM for Multi-POI Extraction from User Queries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems using AI and NLU struggle to identify and extract multiple points of interest from user utterances, as they primarily focus on user intent and fail to capture the central idea or context within natural language sentences.
Innovation Solution
A method employing a bidirectional Long Short Term Memory (LSTM) neural network processes input vectors comprising Part of Speech (POS) vectors, word embeddings, and dependency labels to assign Point of Interest (POI) tags, enabling the extraction of POI text from user queries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional AI and NLU systems focus on identifying user intent, then user intent identification is achieved, but multiple points of interest in a single user utterance cannot be identified and extracted
Solution Approach 1:
The system segments the user utterance into multiple target words and processes each word independently through the bidirectional LSTM neural network. Each word is assigned POI tags separately, allowing multiple points of interest to be identified within a single utterance. This segmentation approach enables the system to extract multiple POIs that were previously missed by conventional intent-only identification methods.
2Productivity
If conventional systems use simple intent identification methods, then processing speed is maintained, but multiple points of interest cannot be captured in user queries
Solution Approach 1:
The system introduces an intermediary bidirectional LSTM neural network layer between the input utterance and the final POI extraction. This intermediary processor analyzes each target word in context by processing input vectors that include the target word, preceding words, POS tags, word embeddings, and dependency labels. The neural network acts as a mediator that captures multiple POIs while maintaining systematic processing.
3Loss of information
If conventional NLU systems process only user intent, then processing efficiency is maintained, but context and central idea of user queries are not captured
Solution Approach 1:
The system adds another dimension to the processing by creating multi-dimensional input vectors for each target word. These vectors incorporate not only the target word itself but also at least two preceding words, POS tags, word embeddings, head word embeddings from dependency parse trees, and dependency labels. This dimensional expansion allows the system to capture contextual information and central ideas that single-dimension intent identification misses.
Data Source
AI summary
A method and device for extracting Point of Interest (POI) from natural language sentences is disclosed. The method includes creating an input vector comprising a plurality of parameters for each target word in a sentence inputted by a user. The method further includes processing for each target word, the input vector through a trained bidirectional LSTM neural network, which is trained to identify POI from a plurality of sentences. The method includes associating POI tags to each target word in the sentence based on processing of associated input vector through the trained bidirectional LSTM neural network. The method further includes extracting POI text from the sentence based on the POI tags associated with each target word in the sentence. The method further includes providing a response to the sentence inputted by the user based on the POI text extracted from the sentence.