Questions and Answers
How can I improve retrieval accuracy for a certain memo?By recording voice-tags several times within a memo, its retrieval accuracy may be improved. Re-recording of voice-tags makes sense, because no word can be uttered in the exact same way, twice. Plus, in the presence of background noise, chances are that different parts of the words get corrupted. Thus, the presence of several variations of the same word increases the probability that for a given query utterance a well matching instance of the sought voice-tag may be found.Why are the memos of a deleted category still present in the category ALL?The original idea behind this product was a comprehensive aide mémoire. Just as from the human brain nothing ever really gets lost under normal circumstances, this category was meant to play the role of the "subconcious" from where things long believed forgotten could be re-collected. Therefore all memos regardless of their actual category are also stored in the category ALL. However, if a memo itself is deleted, it is also removed from the category ALL.Why does the result list contain entries that have nothing to do with what was looked for?Speech processing is not hard logic, rather it falls in the realm of socalled "soft computing". Speech is subject to a certain variability, as man is unable to pronounce a word in exactly the same way, twice. Also, during normal use ideal recording conditions are hardly ever met, so that there's always some background noise in the recordings. Furthermore, words can be so similar to each other that even humans may fail to make a clear decision - this is especially true for telephone speech. Instead of a hard, possibly erroneous decision it is consequently more sensible for a search engine to deliver a "softer" result based on the likeliness of the match. Such relevance oriented result lists are familiar from internet search engines.I went through all entries in the result list, but none contained the word I searched for. Why do I still get a result?As explained in the previous answer, ADnota PMA makes no hard decisions about the equality or inequality of voice tags and queries. Rather it returns a list ordered by decreasing similarity. This means that the first entry of your result is most similar to your query.Is ADnota PMA capable of speech-to-text conversion?No, the current state of the art in speech processing does not allow for speech-to-text conversion on current smartphones. In future releases an interface to a PC based speech-to-text system may be added. However, hopes in such systems shouldn't be set too high. Recording quality and background noise are far below the requirements today's systems pose for good functioning - especially for the hardly constrained topics of daily notes.Why isn't ADnota PMA speech controlled?After our studies of the Series 60 user interface taking a "multimodal" approach seemed ideal to us. The joystick is a quicker and more dependable operating control than voice commands. However, this is not a dogma of ours: Should other platforms or user requests suggest the use of speech control, this can be realised easily with our speech recognition engine in future releases.While inputting a voice-tag a grey bar appears within the red progress bar. What does it mean?The grey bar is a visual feedback on the voice-tag's boundaries. It shows you where in the course of the 2 second recording ADnota PMA detected a suitable signal. If no tag duration bar appears this may indicate that the ambient noise level is too high for the reliable extraction of a voice tag. Due to limitations of update frequency, for some very short yet valid voice-tags no tag duration bar may be shown.After inputting several voice-tags a coloured bar appears. What is it good for?This visual feedback is meant to give the user a feeling for the similarity between two voice-tags. Red color indicates low similarity, whereas yellow or even green colour indicates high similarity.After speaking a voice search query the message "No voice tag detected!" appears. What does it mean?This means that no signal usable for voice search was detected in the query utterance. Probably the environment noise level was too high. As no search is conducted when this message appears, voice search has to be reinitiated. |