I'm building sound recognition model to detect "ding-dong" sound.
There are two procedures, training and testing.
Training data are "ding-dong" sounds generated by a device.
The model can detect "ding-dong" sounds generated by the same device, it works well.
But, when there is a new "ding-dong" sound generated by the second device, the performance will be bad.
I know the possible solution of this issue: record "ding-dong" sound generated by the second device and add it to training data.
But, there is always a new device, new "ding-dong" sound.
What should I do ?
You are facing overfitting problem. Overfitting means that your model has trained to work optimally on specific cases which are the training data set. In order to overcome this problem you should train your model on many devices and then make interpolation between them. Interpolation may be guaranteed by the model you are using.
However, the previous information is so general. In your case, you may find much much easier way to do it. All is depend about how you define "ding-dong". If you could find a siguntur for the "ding-dong" it would be great. This signature should be invariant to all undesirable features.
For example, should "Diiiiing-doooooong" be accepted? if yes, you should find a signature which invariant to length audio clip. Is "ding-dong" with higher frequency acceptable? If yes, you should find a signature which take frequencies as fraction of each other not as absolute values and so on...
BTW, I am sure you may google this and find many many papers about your problem but it may be about "dang-dong" not "ding-dong" but you will still able to benefit from it ;)
So you want to recognize "ding dong sounds" from "other sounds".
One approach could be to train also the data to recognize "other sounds" as an other class. Therefore, a new ding dong could be more easily associated with the "ding-dong sounds" than the "others-sounds".
One drawback of this method could be a growth in the number of "false alarms", but this task always leads to a compromise between precision and recall.
Related
I trained a GPT-J and GPT-Neo models (fine tuning) on my texts and am trying to generate new text. But very often the sentences are very long (sometimes 300 characters each), although in the dataset the sentences are of normal length (50-100 characters usually). I tried a lot of things, changed, adjusted the temperature, top_k, but still half of the results with long phrases and I neen more short.
What can you try?
Here are long examples of generated results:
The support system that they have built has allowed us as users who
are not code programmers or IT administrators some ability to create
our own custom solutions without needing much programming experience
ourselves from scratch!
All it requires are documents about your inventory process but
I've found them helpful as they make sure you do everything right for
maximum efficiency because their knowledge base keeps reminding me
there's new ways i can be doing some things wrong since upgrading my
license so even though its good at finding errors with documentation
like an auditor may bring up later downline someone else might benefit
if those files dont exist anymore after one year when upgrades renews
automatically!
With all GPT models you can specify the "max_length" parameter during generation. This will force the model to generate an amount of tokens equal to max_length. You could also play with num_return_sequences and use a helper function to choose the shortest sequence.
Example:
output = model.generate(input_ids, do_sample=True, top_k=50, max_length=100, top_p=0.95, num_return_sequences=1)
These large language models are trained on massive amounts of data, and fine-tuning them can take patience as they learn to adapt to what you're feeding it. Try different things - adjust your training data format, try different samples, use a pre-prompt during generation to guide the model, etc.. A model like GPT-J does a mind-numbingly large amount of calculations just to spit out a single word, so it is hard to predict what exactly is causing it to say one thing over another.
let me introduce the context briefly: I'm fine tuning a generic BERT model for the context of food and beverage. The final goal is a classification task.
To train this model, I'm using a corpus of text gathered from blog posts, articles, magazines etc... that cover the topic.
I am however facing a predicament that I don't know how to handle: specifically, there are sometimes words that either contain a typo, or maybe different accents, but that are semantically the same.
Let me give you an example to briefly illustrate what I mean:
The wine Gewürztraminer is correctly written with the ü, however sometimes you also find it written with just a normal u, or some other times even just Gewurtz. There are several situations like this one.
Now, a human being would obviously know that we're talking exactly about the same thing, but I have absolutely no idea about how BERT would handle these situations. Would it understand that they're the same thing? Would it consider them instead to be completely different words?
I am currently in the process of cleaning my training data, fixing the typos and trying to even out all these inconsistencies, but at this point I'm not even sure if I should do that at all, considering that the text that will need to be classified can potentially contain typos and situations like the one described above.
What would you guys suggest?
I've been working on a sentence transformation task that involves paraphrase identification as a critical step: if we are confident enough that the state of the program (a sentence repeatedly modified) has become a paraphrase of a target sentence, stop transforming. The overall goal is actually to study potential reasoning in predictive models that can generate language prior to a target sentence. The approach is just one specific way of reaching that goal. Nevertheless, I've become interested in the paraphrase identification task itself, as it's received some boost from language models recently.
The problem I run into is when I manipulate sentences from examples or datasets. For example, in this HuggingFace example, if I negate either sequence or change the subject to Bloomberg, I still get a majority "is paraphrase" prediction. I started going through many examples in the MSRPC training set and negating one sentence in a positive example or making one sentence in a negative example a paraphrase of the other, especially when doing so would be a few word edit. I found to my surprise that various language models, like bert-base-cased-finetuned-mrpc and textattack/roberta-base-MRPC, don't change their confidences much on these sorts of changes. It's surprising as these models claim an f1 score of 0.918+. The dataset is clearly missing a focus on negative examples and small perturbative examples.
My question is, are there datasets, techniques, or models that deal well when given small edits? I know that this is an extremely generic question, much more than is typically asked on StackOverflow, but my concern is in finding practical tools. If there is a theoretical technique, then it might not be suitable as I'm in the category of "available tools define your approach" rather than vice-versa. So I hope that the community would have a recommendation on this.
Short answer to the question: yes, they are overfitting. Most of the important NLP data sets are not actually well-crafted enough to test what they claim to test, and instead test the ability of the model to find subtle (and not-so-subtle) patterns in the data.
The best tool I know for creating data sets that help deal with this is Checklist. The corresponding paper, "Beyond Accuracy: Behavioral Testing of NLP models with CheckList" is very readable and goes into depth on this type of issue. They have a very relevant table... but need some terms:
We prompt users to evaluate each capability with
three different test types (when possible): Minimum Functionality tests, Invariance, and Directional Expectation tests... A Minimum Functionality test (MFT), is a collection of simple examples (and labels) to check a
behavior within a capability. MFTs are similar to
creating small and focused testing datasets, and are
particularly useful for detecting when models use
shortcuts to handle complex inputs without actually
mastering the capability.
...An Invariance test (INV) is when we apply
label-preserving perturbations to inputs and expect
the model prediction to remain the same.
A Directional Expectation test (DIR) is similar,
except that the label is expected to change in a certain way. For example, we expect that sentiment
will not become more positive if we add “You are
lame.” to the end of tweets directed at an airline
(Figure 1C).
I haven't been actively involved in NLG for long, so this answer will be a bit more anecdotal than SO's algorithms would like. Starting with the fact that in my corner of Europe, the general sentiment toward peer review requirements for any kind of NLG project are higher by several orders of magnitude compared to other sciences - and likely not without reason or tensor thereof.
This makes funding a bigger challenge, so wherever you are, I wish you luck on that front. I'm not sure of how big of a deal this site is in the niche, but [Ehud Reiter's Blog][1] is where I would start looking into your tooling ideas.
Maybe even reach out to them/him personally, because I can't think of another source that has an academic background and a strong propensity for practical applications of NLG, at least based on the kind of content they've been putting out over the years.
Your background, environment/funding, and seniority level/control you have over the project will eventually compose your vector decision for you. I's just how it goes on the bleeding edge of anything. What I will add, though, is not to limit yourself to a single language or technology in this phase because of those precise reasons you've mentioned. I'd recommend the same in terms of potential open source involvement but if your profile information is accurate, that probably won't happen, no matter what you do and accomplish.
But yeah, in the grand scheme of things, your question is far from too broad, in my view. It identifies a rather unmistakable problem pattern that not all branches of science are as lackadaisical to approach as NLG-adjacent fields seem to be right now. In that regard, it's not broad enough and will need to be promulgated far and wide before community-driven tooling will give you serious options on a micro level.
Blasphemy, sure, but the performance is already stacked against you As for the question potentially being too broad, I'd posit it is not broad enough, so long as we collectively remain in a "oh, I was waiting for you to start doing something about it" phase.
P.S. I'd eliminate any Rust and ECMAScript alternatives prior to looking into Python, blapshemous as this might sound to a 2021 data scientist
. Some might ARight nowccounting forr the ridicule this would receive xou sltrsfx hsbr s fszs drz zhsz s mrnzsl rcrtvidr, sz lrsdz
due to performance easons.
[1]: https://ehudreiter.com/2016/12/18/nlg-vs-templates/
Is there any way to test multiple algorithms rather than doing it once for each and every algorithm; then checking the result? There are a lot of times where I don’t really know which one to use, so I would like to test multiple and get the result (error rate) fairly quick in Azure Machine Learning Studio.
You could connect the scores of multiple algorithms with an 'Evaluate Model' button to evaluate algorithms against each other.
Hope this helps.
The module you are looking for, is the one called “Cross-Validate Model”. It basically splits whatever comes in from the input-port (dataset) into 10 pieces, then reserves the last piece as the “answer”; and trains the nine other subset models and returns a set of accuracy statistics measured towards the last subset. What you would look at is the column called “Mean absolute error” which is the average error for the trained models. You can connect whatever algorithm you want to one of the ports, and subsequently you will receive the result for that algorithm in particular after you “right-click” the port which gives the score.
After that you can assess which algorithm did the best. And as a pro-tip; you could use the Filter-based-feature selection to actually see which column had a significant impact on the result.
You can check section 6.2.4 of hands-on-lab at GitHub https://github.com/Azure-Readiness/hol-azure-machine-learning/blob/master/006-lab-model-evaluation.md which focuses on the evaluation of multiple algorithms etc.
I have 5 recorded wav files. I want to compare the new incoming recordings with these files and determine which one it resembles most.
In the final product I need to implement it in C++ on Linux, but now I am experimenting in Matlab. I can see FFT plots very easily. But I don't know how to compare them.
How can I compute the similarity of two FFT plots?
Edit: There is only speech in the recordings. Actually, I am trying to identify the response of answering machines of a few telecom companies. It's enough to distinguish two messages "this person can not be reached at the moment" and "this number is not used anymore"
This depends a lot on your definition of "resembles most". Depending on your use case this can be a lot of things. If you just want to compare the bare spectra of the whole file you can just correlate the values returned by the two ffts.
However spectra tend to change a lot when the files get warped in time. To figure out the difference with this, you need to do a windowed fft and compare the spectra for each window. This then defines your difference function you can use in a Dynamic time warping algorithm.
If you need perceptual resemblance an FFT probably does not get you what you need. An MFCC of the recordings is most likely much closer to this problem. Again, you might need to calculate windowed MFCCs instead of MFCCs of the whole recording.
If you have musical recordings again you need completely different aproaches. There is a blog posting that describes how Shazam works, so you might be able to find this on google. Or if you want real musical similarity have a look at this book
EDIT:
The best solution for the problem specified above would be the one described here ("shazam algorithm" as mentioned above).This is however a bit complicated to implement and easier solution might do well enough.
If you know that there are only 5 different different possible incoming files, I would suggest trying first something as easy as doing the euclidian distance between the two signals (in temporal or fourier). It is likely to give you good result.
Edit : So with different possible starts, try doing an autocorrelation and see which file has the higher peak.
I suggest you compute simple sound parameter like fundamental frequency. There are several methods of getting this value - I tried autocorrelation and cepstrum and for voice signals they worked fine. With such function working you can make time-analysis and compare two signals (base - to which you compare, in - which you would like to match) on given interval frequency. Comparing several intervals based on such criteria can tell you which base sample matches the best.
Of course everything depends on what you mean resembles most. To compare function you can introduce other parameters like volume, noise, clicks, pitches...