How it works

Source Data

  1. Our data

    • Social Data

    • Online reviews and blogs, Twitter and etc.

    • Internal data

    • First-Party Data, Additional text data, engagement metrics, etc.

    • Second-Party and Third-Party Data

    • Images, detailed item attributes, distribution channels, price information and etc.

      • Our data is obtained from reliable suppliers

  2. Your data

    • Item metadata

    • Unique ID values for each item, the name of the item, item review data, and etc.

    • User events/ interactions

    • User events/ interaction data includes User Like data, User Engagement data, etc.

    • User metadata

    • User gender, geo-location, etc.

      • Your data will be kept private and secure, and only authorized personnel of our company will be able to view and work with your data. Your data will not be used for any other purpose without your prior consent.

Engineering Process

The essence of our engineering process lies in keeping things simple for you. Keytalk’s engineering process utilizes advanced automation to reduce the time and resources needed to build a good recommendation engine.

The Simple Steps Needed

!theme materia-outline
participant User
participant Item
participant "Keytalk conf"
participant "Keytalk System"
User -> Item: Transmit item library data
Item -> "Keytalk System": Data Preprocessing & Model Training
"Keytalk System" -> "Keytalk System": Data Mapping & Analysis
"Keytalk System" --> Item: Transfer Results
Item --> User: Check Results
User -> "Keytalk conf": Make Configuration Adjustments
"Keytalk conf" --> User: Check Results

We have pre-collected data from three industries: Beauty & Cosmetics, Movies & TV, and Travel. If your business is included in either of these 3 categories, the process should approximately take a month. If your business is not included in the 3 industries, the entire process should take around 3 months to complete.

  1. Transmit Item Library Data:

    Provide us with information on your business goals, and item library data. For data transmission, feel free to use your choice of channel (We’re flexible) Past customers’ methods of communication include cloud storage software, messenger, and e-mail.

  2. Data Preprocessing

    We will proceed to map your item data with user language data. If your site contains real user reviews of items, please share your review data with us for a more relevant Search & Recommendation experience. (The review data you share will not be used for any other purpose than enhancing the relevancy of your DSCM system, and will not be shared with other parties nor used for other projects).

  3. Model Training

    We will begin basic training and customization of the Keytalk ML model. During this step, we will conduct a/b tests of different algorithms and filtering techniques to find the best one that fits your business goals. Once a superior set of methods is found, semantics to use as new search & recommendation signals will be calculated and then refined. The estimated time required to complete this step will vary on factors such as the number of items in your library and desired business goals.

  4. Transfer Results

    Once the Keytalk model is deployed, we will transmit the results with an API.

  5. Check Results

    If you run into any counterintuitive or seemingly implausible results while reviewing your received data, do not worry. This may happen because AI is essentially a machine that understands everything through only numbers and scores, whereas humans achieve understanding on a more collective foundation. As a solution, you can log on to our Solution Admin page to make any necessary adjustments to AI-generated results. Click here to learn more about the functions and uses of our Solution Admin.