Identifying the relevant Entity name requires the ability to:
- Understand the Name in Multiple Languages. Example: Khaled Mashal in arabic: الد مشعل
- Resolve different Transliterations and Spellings. Example: Khaled Mashal can be spelled according to different pronunciations with either K, Q or H as it’s first letter: Haled, Qaled, Khaled..
- Distinguish between different Name-structures, especially in unstructured names. Example: Unstructured names like Fernando Luis Gomez
- Spelling mistakes. Example: heled mashal, qalid misaall, haled masha1l..
AVOID ETHNIC DISCRIMINATION in NAMES SEARCH
Example: US citizen named – Mohamad.
Simple database search will result with many people named Mohamad;
However, due to the inherent weakness in non-phonetic-based-technologies, it will result in hundreds of False Positives, that can be seen as discrimination and racism against Muslims, in which the name Mohamad is common.
FIND THE RIGHT ENTITY WITHOUT DISCRIMINATION
Therefore, the Key-Challenge is searching in a WIDE-FUNNEL on the one hand, whilst being very ACCURATE in the name matching or entity matching.
Basing this task on human-manual process alone will mean:
- Will have to check each name by a team that covers many different languages
- They will have to go through enormous amounts of data
- Each name matching verification will take a relatively long-time
- Bias based on race and nationality will become a major parameter