Name Matching Makes Hotel Self Check-In Possible

Hotels Are Moving to Increased Automation Via Fully Automated Check-In

Hotels are seeing great benefits in moving towards fully automated self check-in procedures. In view of Covid, it makes sense to reduce person-to-person contact. Automating check-in also frees up staff resources for more specialized customer-facing tasks and can reduce operational costs. It can even speed up check-in time, particularly during peak times or group arrivals when the front desk is busy and sufficient staff may not be available.

In order to move to fully automated check-in, the following functions have to be handled:

• Verification of identification documents against reservations
• Dispensing of RFID keys (or keys and codes as alternatives)
• Collection of card or cash payments, including support for different currencies and returning of change in local currencies
• Printing out of receipts

This blog will focus on the first step, identity verification, and how Fuzzy Name Matching, a machine learning technology, can perform it effectively.

Why Fuzzy Name Matching is Critical

A critical step in achieving automation of the check-in procedure is to reliably match information such as person names on identification documents, including driver’s licenses and passports, against that on reservations and possibly also against databases of previous/loyalty customers.

Matching of the names on identification documents to names on reservations can be challenging. First of all, since the identification documents are sometimes scanned and OCR’ed and these processes produce some errors, hotels will need a technology that will still be able to effectively match. There is one available: Fuzzy Name Matching.

In addition to handling simple OCR errors, person name variants also have to be covered. In particular, items such as driver’s licenses or passports will typically have the full form of a name, while a customer may enter their name on the reservation form with a nickname or some other variant, e.g., John Smith vs Johnny Smith. Alternatively, a driver’s license may contain a middle name or a middle initial, while the name on the reservation doesn’t: Mary Catherine Mason vs. Mary C. Mason vs. Mary Mason.

These are the simple cases. Many other features of person names offer more significant challenges to matching. Here’s a sampling from a very long list:

• Names in languages that are written in non-Latin scripts can be transcribed into Latin script in different ways. There is rarely a universally agreed-upon standard for transliteration into Latin script. For example, the Arabic name Abdel Fattah el-Sisi can also be written Abdul Fatah al-Sisi.
• Also, due to the great ethnic diversity in the U.S. and increasingly globally, there will be names of many different ethnic types which have their own unique variations. Here are just a few of them:
  • Spanish names frequently contain both patronymics and matronymics, that is, surnames from the father and mother, respectively. When a full name is shortened, the matronymic is typically dropped. Thus, if a hotel guest checks in as Juan Ramos, a record for Juan Ramos Ramirez would be a better match than a record for Juan Ramirez Ramos.
  • Foreign languages in Latin script also frequently contain diacritics, which may be absent: Joaquín Guzmán vs. Joaquin Guzman
  • Arabic names, which commonly contain an element “al-,” may appear with or without it, as in Musa al-Rashid vs. Musa Rashid
  • Some East Asian names are in the family name + given name order while others are not: Xi Jinping, Moon Jae-in, but Shinzo Abe

Intelligent fuzzy name matching is required to handle all these phenomena. (For more information on the challenges of name matching, see here).

How Fuzzy Name Matching Works

Traditional fuzzy name matching uses combinations of techniques such as Soundex, Metaphone, Edit Distance, and variations of these. More recently, advanced name matching utilizes Machine Learning to train on extensive real-world data consisting of name variants and generate name matching models. This approach automatically learns a collection of intelligent name matching rules from the data. Consequently, these rules reflect countless name variants that occur in the real world and are far more robust and accurate than traditional methods.

Some Fuzzy Name Matching tools also have automated name ethnicity detection capabilities and applies the most appropriate matching models to names based on their ethnicity values in order to attain high accuracy. In addition, these models generate similarity scores that can be used to set thresholds for deciding when a match is close enough for acceptance as a good match. Finally, Fuzzy Name Matching can provide the real-time response times required in a self-check-in scenario.

Summary

In sum, Fuzzy Name Matching is an effective technology using the most advanced Machine Learning techniques to make a critical contribution to fully automated check-in. It will help achieve significant operational efficiencies and increase customer satisfaction in hotel operations.