A recent customer success story made us very happy, and proud of our work. We were working through an evaluation of fuzzy matching individual names with multiple variations – typos, field swapping, case differences, missing parts of the name etc. We trained our machine learning record matching model
on the customer data in about an hour, and got some real good results to highlight. Just as we got ready to present the results, someone in the client team asked us if we could process Hindi information as well. We knew we had done Thai, Japanese, Chinese, French and German so far, but we had never tried Reifier on Hindi data. We knew our AI algorithms are language agnostic, so we thought it would be a good test and validation for us. We ran the model of English names data on Hindi names, and got outstanding results. Wow, no need to retrain! What a pat on our backs ! But, well, evaluations will be evaluations..intrigued, the customer asked us if we could do Hindi addresses. Of course we could, but we had just 45 minutes before the presentation, and we still had not seen their data. Pushing the envelope, we huddled with their database team to let us access their Hindi address data, and ran the “same model trained on English names” on Hindi addresses. With beating hearts and nervous hands, we reviewed the results and were ecstatic to see great quality matches thrown up by Reifier. What a day!
Posted on April 6th, 2017