- The **Beam Search Decoder** for DS2 CTC-trained network follows the similar approach in \[[3](#references)\] with a modification for the ambiguous part, as shown in Figure 2.
- The **Beam Search Decoder** for DS2 CTC-trained network follows the similar approach in \[[3](#references)\] as shown in Figure 2, with two important modifications for the ambiguous parts:
- An **external defined scorer** would be passed into the decoder to evaluate a candidate prefix during decoding whenever a space character appended.
- 1) in the iterative computation of probabilities, the assignment operation is changed to accumulation for one prefix may comes from different paths;
- Such scorer is a unified class, may consisting of language model, word count or any customed evaluators.
- 2) the if condition ```if l^+ not in A_prev then``` after probabilities' computation is deprecated for it is hard to understand and seems unnecessary.
- The **language model** is built from Task 5, with a parameter should be carefully tuned to achieve minimum WER/CER (c.f. Task 7)
- An **external scorer** would be passed into the decoder to evaluate a candidate prefix during decoding whenever a white space appended in English decoding and any character appended in Mandarin decoding.
- Such external scorer consists of language model, word count or any other customed scorers.
- The **language model** is built from Task 5, with parameters should be carefully tuned to achieve minimum WER/CER (c.f. Task 7)
- This decoder needs to perform with **high efficiency** for the convenience of parameters tuning and speech recognition in reality.
- This decoder needs to perform with **high efficiency** for the convenience of parameters tuning and speech recognition in reality.
