metif12/LtRMathIR

Codes for SIGIR 2021 paper Learning to Rank for Mathematical Formula Retrieval

LtRMathIR

Codes for SIGIR 2021 paper Learning to Rank for Mathematical Formula Retrieval

Create Instance

We need formula instances to train our model and also for re-ranking. If used for training it is important to specify the query, formula instance id and relevance score

Feature Computation

After the instances are determine, the next step is to computer the similarity features. This is done in Feature_computation directory

Create SVM-Rank Sample File

Once the similarity measures are calculated, we have to provide sample file to use for SVM-Rank. This done by running create_sample_file.py

Train Model

To train svm-rank model, run train_model.py. Please make sure you have SVM_rank already installed.

Re-rank Results

After SVM-rank model is trained, use rerank_result.py to rerank you retrieval result.