MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202641099853 A) filed by Madhankumar C; Ms Lakshmi B - Rathinam Technical Campus; Yamini M K - Rathinam Technical Campus; Samhitha Harini V - Rathinam Technical Campus; and Harini H - Rathinam Technical Campus on August 19, 2026, for Multi-Source Candidate Data Transformer.
Inventors include Ms Lakshmi B - Rathinam Technical Campus; Yamini M K - Rathinam Technical Campus; Samhitha Harini V - Rathinam Technical Campus; and Harini H - Rathinam Technical Campus.
The application for the patent was published on August 28, 2026, under issue no. 35/2026.
Abstract: Multi-Source Candidate Data Transformer ABSTRACT Recruitment organizations frequently collect candidate information from heterogeneous sources such as resume PDFs, recruiter-maintained spreadsheets, and configuration files, resulting in duplicated records, inconsistent formats, missing attributes, and difficulties in maintaining data traceability. This work proposes a Multi-Source Candidate Data Transformer, a modular and configurable data-processing framework designed to consolidate fragmented candidate information into a standardized and validated representation. The proposed system employs independent source-specific extraction modules for processing resume PDFs, CSV files, and JSON configuration data, followed by a unified pipeline comprising data normalization, duplicate detection, candidate matching, record merging, confidence scoring, provenance tracking, field projection, schema validation, and final JSON generation. During normalization, phone numbers are converted into standardized E.164 format, dates are transformed into a consistent month-year representation, and candidate skills are mapped to canonical skill names using predefined aliases and fuzzy string matching. Candidate identity resolution primarily utilizes email addresses and phone numbers, while fuzzy name matching is applied when unique identifiers are unavailable. Conflicting attributes are resolved using configurable source-priority rules, while confidence scores are calculated based on source reliability, extraction quality, and normalization success. Provenance metadata is associated with individual output fields to provide traceability regarding their source, extraction method, and confidence level. A runtime JSON configuration enables users to customize output fields, confidence and provenance information, and missing-value handling. JSON Schema validation is further employed to ensure structural consistency and integrity of the generated records. Implemented using Python and supporting technologies including Streamlit, Pandas, pdfplumber, BeautifulSoup, RapidFuzz, and phonenumbers, the proposed framework provides a practical, extensible, and configurable solution for improving the accuracy, consistency, traceability, and usability of fragmented recruitment data.
Disclaimer: Curated by HT Syndication.
