Probabilistic Topic Model for Song Recommendation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for matching songs with concepts or moods are computationally expensive and do not accurately correlate songs with particular concepts or moods, relying on signal processing or metadata analysis.

Innovation Solution

A probabilistic topic model is generated from song lyrics, allowing users to submit terms to identify associated topics and recommend songs based on determined probabilities, reducing computational expense and improving accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If signal processing approach is used to analyze acoustic properties of songs, then matching accuracy between songs and concepts is improved, but computational expense increases

Engineering Contradiction:

Improvematching accuracyVSAvoidcomputational expense

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent extracts and utilizes only the textual content (lyrics) from songs, separating it from the full audio signal. This extraction allows for concept matching using simple text processing techniques instead of computationally intensive acoustic analysis, thereby reducing computational expense while maintaining matching accuracy through direct semantic comparison of lyrics with concept descriptions

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent replaces the mechanical signal processing system with a text-based information processing system. Instead of using acoustic feature extraction and signal analysis, the invention uses lyric text processing with concept frequency counting and comparison, substituting complex mechanical computation with simpler information retrieval and text matching operations

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Use of energy by moving object

If metadata-based approach is used to analyze genre and annotations, then computational expense is reduced, but matching accuracy between songs and concepts deteriorates

Engineering Contradiction:

Improvecomputational expenseVSAvoidmatching accuracy

Core Design Contradiction:

Use of energy by moving objectVSMeasurement precision

Solution Approach 1:

The patent performs preliminary action by pre-processing and storing the lyrical content of songs in a searchable format before matching is needed. The lyrics are extracted, tokenized, and prepared for rapid querying, so that when concept matching is required, the system can quickly search and compare without needing to perform complex real-time analysis, thus achieving both low computational expense and high accuracy

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent creates a text-based copy of the song content (lyrics) that can be independently processed and searched. This textual representation serves as a surrogate for the full audio experience, allowing accurate concept matching through text search and comparison without requiring the original audio or complex metadata, thereby reducing computational expense while maintaining matching precision

Inventive Principle:

Principle #26Copying

3Measurement precision

If comprehensive lyric analysis is performed for all songs, then concept matching accuracy is improved, but processing time increases

Engineering Contradiction:

Improveconcept matching accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies partial action by analyzing only the portions of song data that are relevant to concept matching - specifically the lyrical content containing conceptual terms. Rather than performing comprehensive analysis of all song attributes, the system focuses selectively on lyric text that contains or relates to the queried concepts, reducing processing time while maintaining matching accuracy through targeted analysis

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS10452710B2Selecting content items based on received term using topic model

Publication Date: 2019.10.22 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

Lyrics associated with songs are processed to generate a probabilistic topic model that includes probabilities for terms of the lyrics with respect to one or more predetermined topics. At a later time, a user may desire to hear songs that are associated with a particular term, and may submit the term using a user interface. When the term is received, the probabilities of the probabilistic model are used to identify a topic of the predetermined topics that is most likely associated with the received term. The probabilistic model is used to identify songs that are associated with the identified topic, and some or all of the identified songs are presented as being related to the received term in the user interface.