Calculating N-dimensional sentiments using large language models

Abstract

A method for generating predictions related to a set of input text is provided. Text may be received. A machine learning model may be utilized to determine a set of classification probabilities of the text relative to a set of anchors. An emotional score indicative of emotional content of the text may be determined based at least in part on a convex combination of the set of probabilities for the text. One or more predictions related to the text may be generated.

Description

[0001] background

[0002] CROSS-REFERENCE TO RELATED APPLICATIONS

[0003] This application claims priority to and the benefit of U.S. Provisional Application Serial No. 63 / 506,447, filed on June 6, 2023, entitled “FROM ALARGELANGUAGE MODEL TO THREE-DIMENSIONAL SENTIMENT”, U.S. Provisional Application Serial No. 63 / 392,451, filed on July 26, 2022, entitled “PSILOCYBIN THERAPY FOR TREATMENT RESISTANT DEPRESSION”, U.S. Provisional Application Serial No. 63 / 414,769, filed on October 10, 2022, entitled “PSILOCYBIN THERAPY FOR TREATMENT RESISTANT DEPRESSION”, and U.S. Provisional Application Serial No. 63 / 414,769, filed on October 10, 2022, entitled “PSILOCYBIN THERAPY FOR TREATMENT RESISTANT DEPRESSION”, and U.S. Provisional Application Serial No. 63 / 414,769, filed on October 24, 2023, entitled “PSILOCYBIN THERAPY FOR TREATMENT No. PCT / US23 / 070857, the entire disclosure of which is hereby incorporated by reference in its entirety for all purposes.

[0004] Recent advances in artificial intelligence have provided the necessary building blocks for quantifying emotion in human language. In the context of mental health, an important form of human interaction is that between therapist and patient. The therapeutic administration of psychedelics has shown great potential in both historical records and recent clinical trials for the treatment of depression and other related mental and behavioral disorders. For example, recent studies have shown promising results when psilocybin preparations are used to treat patients with treatment-resistant depression (TRD). However, while promising, this treatment approach may only be effective for a subset of the population, and early prediction of outcomes is a key goal for treatment. Using natural language processing (NLP) methods in conjunction with other approaches to quantify the language spoken during therapeutic sessions at scale can help bring unprecedented precision and rigor to the analysis of human language.

[0005] The psychological description of emotions generally involves one of two approaches: a categorical approach or a dimensional approach. A categorical approach can classify emotions into a number of discrete categories with different boundaries that separate emotional states. In contrast, a dimensional approach to emotions attempts to define a continuous emotion space in which traditional categorical emotions can be placed. While simple NLP methods based on a categorical approach to emotions (such as a positive-negative sentiment classifier) ​​are already available, NLP models based on an n-dimensional approach to emotions can provide a more complete and detailed measurement of language. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Various embodiments according to the present disclosure will be described with reference to the accompanying drawings, in which:

[0007] Figure 1 An example diagram showing anchor points that may be utilized in accordance with various implementations is shown.

[0008] Figure 2A and Figure 2B An exemplary therapy session transcript and a time series showing smoothed averages of the valence scores, arousal scores, and confidence scores of the therapist and patient, respectively, in the session are shown. Figure 2C Shown are example distributions of patients and therapists that may be utilized in accordance with various implementations.

[0009] Figure 3 Exemplary methods are shown that can be utilized to implement one or more aspects of various embodiments.

[0010] Figure 4 Exemplary methods are shown that can be utilized to implement one or more aspects of various embodiments.

[0011] Figure 5 Components of an exemplary computing device are shown that can be utilized in accordance with various implementations.

[0012] Figure 6 An example of an environment for implementing aspects according to various embodiments is shown.

[0013] Figure 7 Components of another exemplary environment are shown in which aspects of various embodiments may be implemented. DETAILED DESCRIPTION

[0014] A three-dimensional valence-arousal-confidence (VAC) framework can be used to describe a wide variety of emotional states representing the full range of human responses, and the three-dimensional valence-arousal-confidence (VAC) framework captures the essential features of categorical emotions. According to various embodiments, a VAC model can be used to express the sentiment of a given input text as a convex combination whose weights are obtained from the classification scores of a zero-shot classifier built from a large language model (LLM). Methods using the VAC model can help to more accurately capture sentiment differences between sentences that were previously missed or misinterpreted by methods that only used categorical methods or low-dimensional models.

[0015] In the following description, various embodiments will be described. For the purpose of explanation, specific configurations and details are set forth to provide a thorough understanding of the embodiments. However, those skilled in the art will also appreciate that the embodiments may be practiced without the specific details. In addition, well-known features may be omitted or simplified to avoid obscuring the embodiments being described.

[0016] Major depressive disorder (MDD) is a debilitating disease and can affect one in six adults for their entire lives. MDD can be characterized by at least one depressive episode that lasts for at least two weeks, and involves significant changes in mood, cognition, and the ability to experience happiness. Although psychotherapy and / or medication can be used to effectively control MDD, some MDD patients, despite having tried multiple treatments, may not respond to treatment. Such individuals may be referred to as patients with refractory depression (TRD). There are a variety of existing options for treatment, but these options are generally determined to be unsatisfactory. Therefore, it is necessary to develop an alternative therapeutic option with improved efficacy for TRD patients. In addition, the acceptability of the condition or treatment may still be an important challenge for such patients.

[0017] Psilocybin is a tryptamine alkaloid whose potential as an effective antidepressant has been investigated in preliminary studies in patients with life-threatening cancers, MDD, and TRD. In at least one psilocybin study using a 25mg dose of COMP360, a group of participants experienced a reduction in depressive symptoms for up to 12 weeks. While promising, these results suggest that only a subset of the TRD population will experience a lasting psilocybin response. Weeks may have passed, and the opportunity for potential add-on treatment may have been wasted. While this example describes the use of the model in the therapeutic arena, the model could be used for any other application related to the quantification of language, or in applications where the emotion of such language needs to be understood.

[0018] Figure 1An exemplary three-dimensional emotion cube 100 is shown that can be utilized according to various embodiments. In at least some embodiments, the model can utilize a weighted average of nine "anchor points" in the three-dimensional cube to generate values ​​along three emotion dimensions: valence 110, arousal 120, and confidence 130. In the case of n dimensions, where n is any positive integer, there are 2 at the vertices in the n-dimensional cube. n Anchor points and other anchor points (such as the anchor point at the center of the n-dimensional cube). "Confidence" according to one or more embodiments can be defined as the degree to which a text conveys an emotional state of confidence and certainty, rather than a confidence indicator provided by a machine learning classifier. The weights of the weighted average can be calculated based on the numbers generated by the classifier for a set of classifications corresponding to each dimension. This method can help predict whether a participant will respond within a specific week in his treatment timeline. The architecture can be constructed by a geometric operation called convex combination. According to this example, nine points 140 of the cube 100 can be considered, centered at the origin: eight corners (±1, ±1, ±1) and a center 150 located at (0, 0, 0). Although this example describes three dimensions, the n-dimensional model can be used for any number of emotional dimensions. The ability to quantify language in this way (higher than a two-dimensional model) can make the results more accurate than those obtained by human therapists. In addition, different humans may have different views on the meaning of a text, so using methods such as those described herein can help ensure more accurate objective results overall.

[0019] According to an example implementation, audio recordings of therapy sessions can be collected and transcribed into conversational text using natural language processing (NLP) techniques. The transcribed text can then be parsed into individual "utterances" that estimate the emotion of the session for the therapist and patient or participant using a sentiment model. Such a model can generate a valence score, an arousal score, and a confidence score for each utterance.

[0020] In at least some embodiments, the audio can be parsed and labeled as utterances. The audio can be processed using a model to predict the patient's tone. In some embodiments, the audio can be associated with a score indicating the patient's tone. The audio score can be compared to a valence score and an arousal score, or the audio score can be interpreted using the valence score and the arousal score. A zero-shot classifier built on a bidirectional autoregressive transformer (BART) autoencoder and a multi-type natural language inference (MNLI) dataset can be used to calculate the sentiment valence score, the arousal score, and the confidence score.

[0021] For a given text u and anchor point p, we can define a number w for each point p(u) (called the weight of u at p). These numbers can satisfy the following properties:

[0022] w p (u)≥0, and

[0023] ∑ p w p (u) = 1,

[0024] Here, the sum is the sum for all anchor points p.

[0025] Using these weights, the three-dimensional VAC sentiment can be represented as:

[0026]

[0027] This can also be expressed as:

[0028]

[0029] Where p = (p1, p2, p3). The nature of the weights can ensure that the sentiment value is between -1 and 1.

[0030] According to an exemplary embodiment, the quote "Heal yourself, with beautiful love, and always remember...you are the medicine" may have the following anchor weight values ​​for the quote:

[0031] w (1，1，1) =0.204866

[0032] w (1，1，-1) =0.223669

[0033] w (1，-1，1) =0.249116

[0034] w (1，-1，-1) =0.130423

[0035] w (-1，1，1) =0.023099

[0036] w (-1，1，-1) =0.012893

[0037] w (-1，-1，1) =0.103748

[0038] w (-1，-1，-1) =0.011839

[0039] w (0，0，0) =0.040327.

[0040] In this case, the sentiment score can be calculated as:

[0041] Valence (u) = 0.204886*1 + 0.249116*1 + 0.130423*1 + 0.023099*(-1) + 0.012893*(-1) + 0.103748*(-1) + 0.011839*(-1) + 0.040327*0 = 0.684102. Using the same method, arousal (u) = -0.032, and confidence (u) = 0.211. Total: (Valence (u), Arousal (u), Confidence (u)) = (0.684102, -0.031864, 0.210515)

[0042] The weights may vary depending on the number of anchor points, the number of dimensions, and the type of input being processed. According to an exemplary embodiment, a zero-shot classifier is used to calculate the anchor weight w for a text string u and an anchor point p. p (u), this zero-shot classifier is built using a BART autoencoder that has been fine-tuned on the Multi-Type Natural Language Inference (MNLI) dataset. The predictions given by the classifiers can be combined to produce the weights used in the model. The choice of classification itself is a hyperparameter that is difficult to fine-tune. Classification can be used to derive these classifications and group them into "anchor lists".

[0043] According to an exemplary embodiment, the input text may be passed to a BART-MNLI classifier, which includes, but is not limited to, the following categories: acceptance, conformity, aggressiveness, astonishment, awe, comfort, confidence, control, dazzle, provocation, destruction, indifference, discouragement, desire, anger, excitement, weakness, frustration, humiliation, inferiority, joy, loneliness, neutrality, nothingness, panic, indulgence, power, relaxation, resistance, reassurance, seriousness, shelter, gentleness, perseverance, pain, surprise, fear, tolerance, calmness, complacency, victory, and wisdom. Any number of categories may be selected, including more or fewer categories than those described herein.

[0044] For text u, the value P c (u) = the probability that u belongs to category c. This probability can be calculated under the assumption that c belongs to one of the categories and that the categories are mutually exclusive. According to an exemplary embodiment, the categories can be divided into lists using the frequency-inverse document frequency technique, one for each of the nine anchor points p:

[0045] (1, 1, 1)-Category = [excited, joyful, elated, victorious, powerful]

[0046] (1, 1, -1)-category = [amazing, awe, dazzled, eager, surprised]

[0047] (1, -1, 1)-Classification = [Comfortable, Confident, Relaxed, Peace of Mind, Wise]

[0048] (1, -1, -1)-Classification = [Acceptance, Shelter, Gentleness, Calmness, Tolerance]

[0049] (-1, 1, 1)-Category = [aggressive, controlling, provocative, destructive, angry]

[0050] (-1, 1, -1)-Category = [depression, humiliation, panic, pain, fear]

[0051] (-1, -1, 1)-Classification = [Qualified, Indifferent, Resistant, Serious, Perseverant]

[0052] (-1, -1, -1)-Category = [discouraged, weak, inferior, lonely, empty]

[0053] (0,0,0)-Classification = [Neutral, Connivance].

[0054] Weight w p (u) is calculated as:

[0055] w p (u)=∑ c∈p-分类 P c (u).

[0056] The choice of categories and how they are grouped into anchor lists contributes to how the model scores a piece of text. The basic behavior that guides the intuition behind which categories should appear in the anchor list is that the higher a piece of text scores in a category in a given anchor list, the larger the anchor weight will be, and thus the stronger the text will be pulled towards that anchor. Therefore, the categories that appear in the list for a given anchor should reflect the sentiment represented by the anchor.

[0057] For example, if "joy" is in the anchor list for anchor (1, 1, 1), and a piece of text scores high in the "joy" category, then the model pulls the text towards point (1, 1, 1) with a weight that includes the "joy" category score as a summand. In short, the categories in the category list for (1, 1, 1) should be the vocabulary of (1, 1, 1) (or close to it). The same is true for other anchors.

[0058] The classification list can be compiled by collecting sentences spoken in therapy sessions and scoring them. Once scored, a large number of sentences for each affective state corresponding to the anchor point can be collected, and the most characteristic words in each document can be extracted.

[0059] The sentiment score of an utterance can be determined by the weight w p (u), and these weights can depend on the probability P c (u). For a piece of text u, the value P c (u) depends not only on c, but on the entire set of categories. If different categories are used in the anchor list, even if there is no change in the category c, the value P c (u) will also change. On the surface, it seems that all values ​​of P c (u) must be recalculated from scratch, but using the BART-MNLI zero-shot classifier can greatly speed up the calculation of P c (u) regardless of the other categories involved.

[0060] Inside the BART-MNLI zero-shot classifier, a value Lc(u) is calculated that is independent of the other classes. This value can be called the log-odds of the entailment probability, also known as the "log-odds ratio (logit)". The entailment probability is a prediction that a text u is a member of class c without any information about the other classes. Computing the entailment probability is a task that BART-MNLI is fine-tuned upstream of its zero-shot application. The log-odds ratios of all classes c that are planned to be explored and all words and utterances u that will be used when evaluating the model can be calculated in advance. In this way, it is possible to test as many choices of classes as possible, or to test different distributions of the classes used to anchor the list of classes, in a computationally efficient way.

[0061] Figure 2A and Figure 2B An exemplary therapy session transcript 200 and a time series 210 showing smoothed averages of the valence scores, arousal scores, and confidence scores (220, 230, 240) of the therapist and patient, respectively, during the session are shown. Figure 2C An example distribution 250 of patients and therapists that may be utilized in accordance with various implementations is shown.

[0062] In an exemplary therapy session 200, Gloria, a 30-year-old divorced mother of three, discusses her feelings of anxiety and dissatisfaction with her life. Gloria leads the conversation and expresses her feelings in her own words. According to the VAC model, the average emotion of the session is medium-high valence, neutral arousal, and low confidence for both the therapist and the patient. Figure 2C An overall distribution 250 of valence scores, arousal scores, and confidence scores ( 270 , 280 , 290 ) for the patient and therapist in the exemplary therapy session 200 is shown, provided graphically in the form of a probability density 240 .

[0063] Figure 3 An exemplary method 300 that can be utilized to implement one or more aspects of various embodiments is shown. According to an exemplary embodiment, text can be received 310. For example, an audio recording can be generated at the time of an initial integration session between a patient or participant and a medical provider (such as a therapist). Using one or more NLP techniques, the recording can be transcribed into text.

[0064] According to one or more embodiments, a large trained language model can be utilized to help score the utterance.For example, a machine learning model can analyze the text and determine a set of classification probabilities 320 for the text relative to a set of anchor points.

[0065] Text can be scored or valued along three or more dimensions, including but not limited to valence, arousal, and confidence. In this way, sentiment can capture intensity, rather than just positivity or negativity. The sentiment score of a piece of text may be different from trying to infer the speaker's emotional state. For example, a text stating "I love broccoli" may be scored as positive according to traditional sentiment models. However, if voiced in a sarcastic manner, the text will signal a negative attitude towards broccoli.

[0066] A sentiment score indicating the emotional content of the text may be determined 330 based at least in part on a convex combination of a set of probabilities for the text. In certain exemplary embodiments, a classifier may be utilized to classify the text. Such a classifier may be constructed based on a BART autoencoder and a multi-type natural language inference (MNLI) dataset. Using a model built on top of a larger model may enable nearly unlimited free available data to be used for smaller data tasks.

[0067] One or more predictions 340 associated with the text may be generated. For example, the prediction may correspond to a predicted response to an administered treatment, or any other type of prediction associated with a group of text. It should be understood that for any process herein, unless otherwise expressly stated, additional, fewer, or alternative steps may be performed in a similar or alternative order or in parallel within the scope of various embodiments.

[0068] Figure 4An exemplary method 400 that can be utilized to implement one or more aspects of various embodiments is shown. According to an exemplary embodiment, a set of input text can be passed through a BART-MNLI classifier 410. Probability values ​​420 indicating the probability of a text string in the set of input text being within a given classification can be calculated. Grouped classifications 430 for the set of input text can be generated based on the calculated probability values. The grouped classifications can be divided into lists 440 corresponding to one or more anchor points. Weights 450 for the text strings can be generated based at least in part on the lists.

[0069] As explained, the input text may be passed to a BART-MNLI classifier, including but not limited to the following categories: acceptance, conformity, aggressiveness, astonishment, awe, comfort, confidence, control, dazzle, provocation, destruction, indifference, discouragement, desire, anger, excitement, weakness, frustration, humiliation, inferiority, joy, loneliness, neutrality, nothingness, panic, indulgence, power, relaxation, resistance, reassurance, seriousness, shelter, gentleness, perseverance, pain, surprise, fear, tolerance, calmness, triumph, victory, and wisdom. According to an exemplary embodiment, the categories may be divided into lists, one for each of the nine anchor points p:

[0070] (1, 1, 1)-Category = [excited, joyful, elated, victorious, powerful]

[0071] (1, 1, -1)-category = [amazing, awe, dazzled, eager, surprised]

[0072] (1, -1, 1)-Classification = [Comfortable, Confident, Relaxed, Peace of Mind, Wise]

[0073] (1, -1, -1)-Classification = [Acceptance, Shelter, Gentleness, Calmness, Tolerance]

[0074] (-1, 1, 1)-Category = [aggressive, controlling, provocative, destructive, angry]

[0075] (-1, 1, -1)-Category = [depression, humiliation, panic, pain, fear]

[0076] (-1, -1, 1)-Classification = [Qualified, Indifferent, Resistant, Serious, Perseverant]

[0077] (-1, -1, -1)-Category = [discouraged, weak, inferior, lonely, empty]

[0078] (0,0,0)-Classification = [Neutral, Connivance].

[0079] Any number of categories may be selected, including more or fewer categories than described herein.

[0080] Computing resources (such as servers) that can have software and / or firmware updated in this manner will typically include at least a set of standard components configured for general purpose operation, although various proprietary components and configurations may also be used within the scope of various embodiments. Figure 5 Components of an exemplary computing device 500 that can be utilized according to various embodiments are shown. As is known for computing devices, a computer will have one or more processors 502, such as a central processing unit (CPU), a graphics processing unit (GPU), etc., which are electronically and / or communicatively coupled to various components using various buses, traces, and other such mechanisms. The processor 502 may include memory registers 506 and cache memory 504, which are used to store instructions, data, etc. In this example, a chipset 514, which may include a north bridge and a south bridge in some embodiments, may work with various system buses to connect the processor 502 to components such as a system memory 516 in the form of physical RAM or ROM, which may include code for an operating system and various other instructions and data for the operation of the computing device. The computing device may also contain or communicate with one or more storage devices 520 (such as a hard drive, a flash drive, an optical storage, etc.) for persistent storage of data and instructions similar to or in addition to the data and instructions stored in the processor and memory. The processor 502 may also communicate with various other components via the chipset 514 and an interface bus (or graphics bus, etc.), which may include a communication device 524 (such as a cellular modem or network card), a media component 526 (such as a graphics card and audio components), and a peripheral interface 530 for connecting peripheral devices (such as a printer, keyboard, etc.). At least one cooling fan 532 or other such temperature regulating or reducing component may also be included, which may be driven by the processor or triggered by various other sensors or components on the device or remote from the device. Various other or alternative components and configurations for computing devices that are known in the art may be utilized.

[0081] In some embodiments, at least one processor 502 can obtain data from physical memory 516 (such as dynamic random access memory (DRAM) modules) via a consistency architecture. It should be understood that various architectures can be used for such computing devices, which can include different selections, numbers, and parameters of buses and bridges within the scope of various embodiments. The data in the memory can be managed and accessed by a memory controller (such as a DDR controller) through a consistency architecture. In at least some embodiments, the data can be temporarily stored in a processor cache 504. The computing device 500 can also support multiple I / O devices using a set of I / O controllers connected via an I / O bus. There can be I / O controllers for supporting various types of I / O devices, such as universal serial bus (USB) devices, data storage devices (e.g., flash memory or disk storage), network cards, peripheral component interconnect express (PCIe) cards or interfaces 530, communication devices 524, graphics or audio cards 526, and direct memory access (DMA) cards, as well as other such options. In some embodiments, components such as a processor, controller, and cache may be configured on a single card, board, or chip (i.e., a system-on-chip implementation), while in other embodiments at least some of the components may be located in different locations, etc.

[0082] The operating system (OS) running on the processor 502 can help manage various devices that can be used to provide input to be processed. This can include, for example, using relevant device drivers to implement interactions with various I / O devices, where these devices can be related to data storage, device communication, user interface, etc. Various I / O devices will typically be connected via various device ports and communicate with the processor and other device components through one or more buses. There may be a specific type of bus that provides communication according to a specific protocol, such as may include peripheral component interconnect (PCI) or small computer system interface (SCSI) communication, and other such options. Communication can be carried out using registers associated with corresponding ports, including registers such as data input and data output registers. Communication can also be carried out using memory mapped I / O, in which a portion of the processor's address space is mapped to a specific device, and data is directly written to that portion of the address space, and the data is read out from that portion of the address space.

[0083] For example, such a device can be used as a server in a server farm or a data warehouse. A server computer usually needs to perform tasks outside the environment of a CPU and a main memory (i.e., RAM). For example, a server may need to communicate with an external entity (e.g., other servers) or use an external processor (e.g., a general purpose graphics processing unit (GPGPU)) to process data. In this case, the CPU can be docked with one or more I / O devices. In some cases, these I / O devices can be dedicated hardware designed to perform a specific role. For example, an Ethernet network interface controller (NIC) can be implemented as an application specific integrated circuit (ASIC), which includes digital logic that is operable to send and receive packets.

[0084] In an illustrative embodiment, the host computing device is associated with various hardware components, software components, and corresponding configurations that promote the execution of I / O requests. One such component is an I / O adapter that inputs and / or outputs data along a communication channel. In one aspect, the I / O adapter device can communicate as a standard bridge component to facilitate access between various physical and simulated components and a communication channel. In another aspect, the I / O adapter device may include an embedded microprocessor to allow the I / O adapter device to perform computer executable instructions related to the implementation of management functions or the management of one or more such management functions, or to perform other computer executable instructions related to the implementation of the I / O adapter device. In some embodiments, the I / O adapter device can be implemented using multiple discrete hardware elements (such as multiple cards or other devices). The management controller can be configured to be electrically isolated from any other component in the host device except the I / O adapter device. In some embodiments, the I / O adapter device is attached to the host device from the outside. In some embodiments, the I / O adapter device is internally integrated into the host device. Also in communication with the I / O adapter device may be an external communication port component for establishing a communication channel between the host device and one or more network-based services or other network-attached or directly-attached computing devices. Exemplarily, the external communication port component may correspond to a network switch, sometimes referred to as a top-of-rack ("TOR") switch. The I / O adapter device may utilize the external communication port component to maintain a communication channel between one or more services and the host device, such as a health check service, a financial service, and the like.

[0085] The I / O adapter device can also communicate with a basic input / output system (BIOS) component. The BIOS component can include a non-transient executable code, commonly referred to as firmware, which can be executed by one or more processors and is used to initialize the components of the host device and identify system devices, such as video display cards, keyboards and mice, hard drives, optical drives, and other hardware. The BIOS component can also include boot loader software or locate the boot loader software, which will be used to guide the host device. For example, in one embodiment, the BIOS component can include an executable code that, when executed by a processor, causes the host device to attempt to locate the pre-boot execution environment (PXE) boot software. In addition, the BIOS component can include a hardware latch, or the BIOS component utilizes a hardware latch that is electrically controlled by the I / O adapter device. The hardware latch can limit access to one or more aspects of the BIOS component, such as controlling the modification or configuration of the executable code maintained in the BIOS component. The BIOS component can be connected to multiple additional computing device resource components (such as processors, memories, etc.) (or the BIOS component can communicate with multiple additional computing device resource components). In one embodiment, such a computing device resource component can be a physical computing device resource that communicates with other components via a communication channel. The communication channel can correspond to one or more communication buses through which the components of the bare metal host device communicate, such as a shared bus (e.g., a processor bus, a memory bus), a point-to-point bus (such as a PCI or PCI Express bus), etc. Other types of communication channels, communication media, communication buses, or communication protocols (e.g., Ethernet communication protocols) can also be utilized. In addition, in other embodiments, one or more computing device resource components in the computing device resource component can be a virtualized hardware component simulated by the host device. In such an embodiment, an I / O adapter device can implement a management process in which the host device is configured with physical or simulated hardware components based on various standards. The computing device resource component can communicate with the I / O adapter device via a communication channel. In addition, the communication channel can connect the PCI Express device to the CPU via a north bridge or a main bridge, as well as such options.

[0086] One or more controller components may be communicated with the I / O adapter device via a communication channel, and the one or more controller components are used to manage a hard drive or other forms of memory. An example of a controller component may be a SATA hard drive controller. Similar to a BIOS component, a controller component may include a hardware latch or utilize a hardware latch that is electrically controlled by the I / O adapter device. A hardware latch may limit access to one or more aspects of a controller component. Illustratively, the hardware latch may be controlled together or independently. For example, an I / O adapter device may selectively close a hardware latch for one or more components based on a trust level associated with a particular user. In another example, an I / O adapter device may selectively close a hardware latch for one or more components based on a trust level associated with an author or publisher of an executable code to be executed by the I / O adapter device. In another example, an I / O adapter device may selectively close a hardware latch for one or more components based on a trust level associated with the component itself. The host device may also include an additional component that communicates with one or more of the illustrative components associated with the host device. Such components may include devices such as one or more controllers in combination with one or more peripheral devices (such as a hard disk or other storage device). In addition, the additional components of the host device may include another set of peripheral devices, such as a graphics processing unit ("GPU"). The peripheral devices may also be associated with hardware latches to limit access to one or more aspects of the components. As mentioned above, in one embodiment, the hardware latches may be controlled together or independently.

[0087] As discussed, different methods may be implemented in various environments depending on the described implementation. For example, Figure 6An example of an environment 600 for implementing aspects according to various embodiments is shown. As will be appreciated, although a Web-based environment is used for the purpose of explanation, various embodiments may be implemented using different environments as needed. The system includes an electronic client device 602, which may include any appropriate device operable to send and receive requests, messages, or information through an appropriate network 604 and transmit the information back to the user of the device. Examples of such client devices include personal computers, mobile phones, handheld messaging devices, laptops, set-top boxes, personal data assistants, e-book readers, and the like. Examples of such recipients or users may include medical providers (including therapists) or patients. The network may include any appropriate network, including an intranet, an internet network, a cellular network, a local area network, or any other such network or combination thereof. The components used for such a system may depend at least in part on the type of selected network and / or environment. Protocols and components for communicating via such a network are well known, and will not be discussed in detail herein. Communication on the network may be implemented via wired or wireless connections and combinations thereof. In this example, the network includes the Internet because the environment includes a Web server 606 for receiving requests and providing content in response to the requests, but for other networks, alternative devices for similar purposes may be used, as will be apparent to those of ordinary skill in the art.

[0088] The illustrative environment includes at least one application server 608 and data storage device 610. It should be understood that there may be multiple application servers, layers or other elements, processes, or components, which may be linked or otherwise configured, and which may interact to perform tasks such as obtaining data from appropriate data storage devices. As used herein, the term "data storage device" refers to any device or combination of devices capable of storing, accessing and retrieving data, which may include any combination and number of data servers, databases, data storage devices, and data storage media in any standard, distributed, or clustered environment. The application server 608 may include any appropriate hardware and software for integrating with the data storage device 610 as needed to perform aspects of one or more applications for client devices, and for processing most of the data access and business logic for the application. The application server collaborates with the data storage device to provide access control services, and the application server is capable of generating content (such as text, graphics, audio, and / or video) to be transmitted to the user, which may be provided to the user by the Web server 606 in the form of HTML, XML, or other appropriately structured languages ​​in this example. The processing of all requests and responses, as well as the delivery of content between the client device 602 and the application server 608, can be handled by the web server 606. It should be understood that the web server and application server are not required and are merely example components, as the structured code discussed herein can be executed on any appropriate device or host machine as discussed elsewhere herein.

[0089] The data storage device 610 may include a plurality of separate data tables, databases, or other data storage mechanisms and media for storing data related to a particular aspect. For example, the data storage device shown includes a mechanism for storing biomarker data (e.g., production data) 612 and user information 616, which can be used to provide content to the production party. The data storage device is also shown as including a mechanism for storing log or conversation data 614. It should be understood that there may be many other aspects that may need to be stored in the data storage device (such as page image information and access rights information), which can be stored in any of the mechanisms listed above or in additional mechanisms in the data storage device 610 as needed. The data storage device 610 can be operated by logic associated with it to receive instructions from the application server 608 and respond to the instructions to obtain data, update data, or otherwise process data. In one example, a user can submit a search request for a certain type of item. In this case, the data storage device can access the user information to verify the identity of the user, and the data storage device can access the directory details to obtain information about the type of item. The information, including biomarkers and diagnostic data, may then be returned to the user, such as through a patient or therapist portal, accessible via a web page that the user may view via a browser on the user device 602. Information for a particular item of interest may be viewed in a dedicated page or window in the browser.

[0090] Each server will typically include an operating system that provides executable program instructions for the general management and operation of the server, and each server will typically include a computer-readable medium storing instructions that, when executed by a processor of the server, allow the server to perform its intended functions. Suitable implementations for the operating system and the general functionality of the servers are known or commercially available and are readily implemented by those of ordinary skill in the art, particularly in light of the disclosure herein.

[0091] In one embodiment, the environment is a distributed computing environment utilizing multiple computer systems and components that are interconnected via communication links, using one or more computer networks, or direct connections. However, one of ordinary skill in the art will recognize that such a system may be implemented in environments with more than Figure 6 The system may operate equally well in systems having fewer or more components than those shown in FIG. Figure 6 The description of system 600 in FIG. 6 should be considered illustrative rather than limiting the scope of the present disclosure.

[0092] Figure 7An example environment 700 in which aspects of various embodiments can be implemented is shown. In this example, a user can submit a request to a multi-tenant resource provider environment 706 using a client device 702 through at least one network 704. The client device can include any appropriate electronic device that is operable to send and receive requests, messages, or other such information through an appropriate network and communicate the information to the user of the device. Examples of such client devices include personal computers, tablet computers, smart phones, laptops, etc. At least one network 704 may include any appropriate network, including an intranet, an internet network, a cellular network, a local area network (LAN), or any other such network or combination, and communication on the network may be achieved through a wired connection and / or a wireless connection. The resource provider environment 706 may include any appropriate component for receiving requests and returning information or performing actions in response to these requests. For example, the provider environment may include a Web server and / or an application server for receiving requests and processing the requests, and then returning data, web pages, videos, audio, or other such content or information in response to the requests.

[0093] In various embodiments, the provider environment may include various types of resources that can be utilized by multiple users for various purposes. As used herein, computing and other electronic resources utilized in a network environment may be referred to as "network resources". These resources may include, for example, servers, databases, load balancers, routers, etc., which may perform tasks such as receiving data and / or executable instructions, transmitting data and / or executable instructions, and / or processing data and / or executable instructions. In at least some embodiments, all or part of a given resource or group of resources may be allocated to a specific user or allocated for a specific task within at least a certain time period. Sharing these multi-tenant resources from the provider environment is generally referred to as resource sharing, Web services, or "cloud computing," and other such terms, and this depends on a specific environment and / or implementation. In this example, the provider environment includes multiple resources 714 of one or more types. These types may include, for example, an application server or a database server, which may be operable to process instructions provided by a user, and which may be operable to process data stored in one or more data storage devices 716 in response to a user request. As is well known, for such purpose, a user may also reserve at least a portion of the data stored in a given data storage device. Methods for enabling a user to reserve various resources and resource instances are well known in the art, so that a detailed description of the entire process and an explanation of all possible components will not be discussed in detail herein.

[0094] In at least some embodiments, a user who wants to utilize a portion of a resource 714 can submit a request that is received to an interface layer 708 of a provider environment 706. The interface layer may include an application programming interface (API) or other public interface to enable a user to submit a request to a provider environment. The interface layer 808 in this example may also include other components, such as at least one web server, a routing component, a load balancer, and the like. When the interface layer 708 receives a request for providing a resource, information for the request may be directed to a service manager 710 or other such systems, services, or components that are configured to manage user accounts and information, resource supply and use, and other such aspects. The service manager 710 that receives the request may perform tasks such as authenticating the identity of the user submitting the request, and determining whether the user has an existing account at a resource provider, wherein the account data may be stored in at least one account data storage device 712 in the provider environment. A user may provide any of a variety of types of credentials to authenticate the identity of the user to the provider. These credentials may include, for example, a username and password pair, biometric data, a digital signature, or other such information. The provider may verify this information based on the information stored for the user. If the user has an account with appropriate permissions, status, etc., the resource manager can determine whether there are sufficient resources to satisfy the user's request, and if so, can provision resources or otherwise grant the user access to corresponding portions of those resources for the amount specified by the request. The amount can include, for example, the capacity to process a single request or perform a single task, a specified time period, or a recurring / renewable time period, and other such values. If the user does not have a valid account with the provider, the user's account cannot access the resource type specified in the request, or other such reasons are preventing the user from obtaining access to such resources, a communication can be sent to the user to enable the user to create or modify an account, or change the resources specified in the request, and other such options.

[0095] Once the user is authenticated, the account is verified, and the resources are allocated, the user may utilize the allocated resources to achieve a specified capacity, data transfer amount, time period, or other such value. In at least some embodiments, the user may provide a session token or other such credential in subsequent requests to enable such requests to be processed on the user session. The user may receive a resource identifier, a specific address, or other such information that enables the client device 702 to communicate with the allocated resources without having to communicate with the service manager 710, at least until relevant aspects of the user's account change, the user is no longer granted access to the resources, or other such changes occur.

[0096] The service manager 710 (or other such systems or services) in this example can also function as a virtual layer for hardware and software components, processing control functions in addition to management actions, such as supply, expansion, replication, etc. The resource manager can utilize the dedicated APIs in the interface layer 808, each of which can be provided to receive requests for at least one specific action performed relative to the data environment, such as supply, expansion, cloning, or dormant instances. Upon receiving a request for an API in the API, the Web service portion of the interface layer can parse or otherwise analyze the request to determine the steps or actions required to operate or process the call. For example, a Web service call including a request to create a data repository may be received.

[0097] The interface layer 708 in at least one embodiment includes a set of scalable user-oriented servers that can provide various APIs and return appropriate responses based on the API specifications. The interface layer may also include at least one API service layer, which, in one embodiment, includes a stateless replication server that processes external user APIs. The interface layer may be responsible for Web service front-end functions, such as authenticating users based on credentials, authorizing users, limiting user requests to API servers, verifying user input, and marshaling or unmarshaling requests and responses. The API layer may also be responsible for writing database configuration data to the management data storage device in response to API calls, and reading database configuration data from the management data storage device. In many embodiments, the Web service layer and / or the API service layer will be the only externally visible component, or the only component visible and accessible to users of the control service. The servers of the Web service layer may be stateless and may be horizontally scalable, as known in the art. For example, API servers and persistent data storage devices may be spread across multiple data centers in a region, so that the server can withstand single data center failures.

[0098] Various embodiments can also be realized in various operating environments, and in some cases, these operating environments can include one or more user computers or computing devices, which can be used to operate any application in many applications.Users or client devices can include any of many general personal computers, such as desktop computers or notebook computers running standard operating systems, and cellular, wireless and handheld devices that run mobile software and can support many networks and message transfer protocols.Such a system can also include many workstations, which run any of various commercial operating systems and other known applications, for purposes such as development and database management.These devices can also include other electronic devices, such as virtual terminals, thin clients, game systems and other devices that can communicate via a network.

[0099] Most embodiments utilize at least one network familiar to those skilled in the art to support communication using any of a variety of commercially available protocols such as TCP / IP, FTP, UPnP, NFS, and CIFS. The network can be, for example, a local area network, a wide area network, a virtual private network, an internet, an intranet, an extranet, a public switched telephone network, an infrared network, a wireless network, and any combination thereof. In embodiments utilizing a Web server, the Web server can run any of a variety of server or middle-tier applications, including HTTP servers, FTP servers, CGI servers, data servers, Java servers, and business application servers. One or more servers may also be capable of executing programs or scripts in response to requests from user devices, such as by executing one or more Web applications, which may be implemented in any programming language such as Java, PHP, or other programming languages. C, C#, or C++) or any scripting language (such as Perl, Python, or TCL) and combinations thereof. The one or more servers may also include database servers, including but not limited to those available from and etc. A commercially available database server.

[0100] As discussed above, the environment may include various data storage devices and other memories and storage media. These may reside in various locations, such as a storage medium located locally (and / or residing therein) in one or more computers in the computer, or a storage medium remote from any or all computers in the computer via a network. In a specific set of embodiments, information may reside in a storage area network (SAN) familiar to those skilled in the art. Similarly, any file required for performing the functions possessed by a computer, server, or other network device may be stored locally and / or remotely as required. When the system includes a computerized device, each such device may include hardware elements that may be electrically connected via a bus, including, for example, at least one central processing unit (CPU), at least one input device (e.g., a mouse, keyboard, controller, touch-sensitive display element, or keypad), and at least one output device (e.g., a display device, printer, or speaker). Such a system may also include one or more storage devices and removable media devices, memory cards, flash memory cards, etc., such as disk drives, optical storage devices, and solid-state storage devices, such as random access memory (RAM), or read-only memory (ROM). As described above, such devices may also include a computer-readable storage medium reader, a communication device (e.g., a modem, a network card (wireless or wired), an infrared communication device), and a working memory. The computer-readable storage medium reader may be connected to a computer-readable storage medium representing a remote, local, fixed and / or removable storage device and a storage medium for temporarily and / or more permanently containing, storing, transmitting and retrieving computer-readable information, or the computer-readable storage medium reader may be configured to receive a computer-readable storage medium.

[0101] The system and various devices will typically also include multiple software applications, modules, services or other elements located in at least one working memory device, including operating systems and applications, such as client applications or Web browsers. It should be understood that alternative embodiments can have multiple variations different from the embodiments described above. For example, customized hardware and / or specific elements can also be implemented in the form of hardware, software (including portable software, such as applets) or both. In addition, connections with other computing devices (such as network input / output devices) can be adopted. Storage media and other non-transient computer-readable media for containing code or part of code can include any suitable media known or used in the art, such as but not limited to volatile media and non-volatile media, removable media and non-removable media for storing information such as computer-readable instructions, data structures, program modules, or other data implemented in any method or technology, including RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile discs (DVD) or other optical storage devices, cassettes, magnetic tapes, disk storage devices, or other magnetic storage devices, or any other media that can be used to store required information and can be accessed by system devices. Based on the disclosure and teachings provided herein, a person of ordinary skill in the art will appreciate other ways and / or methods to implement the various embodiments.

[0102] Accordingly, the specification and drawings are to be regarded in an illustrative rather than a restrictive sense. It will, however, be evident that various modifications and changes may be made thereto without departing from the broader spirit and scope of the invention as set forth in the claims.

Claims

1. A computer-implemented method, the computer-implemented method comprising: Receive text; using a machine learning model to determine a set of classification probabilities for the text relative to a set of anchor points; as well as A sentiment score indicative of emotional content of the text is determined using the set of probabilities for the text based at least in part on a convex combination of the set of anchor points.

2. The computer-implemented method of claim 1 , wherein: The classification probabilities are calculated using the output of a large language model.

3. The computer-implemented method of claim 2, wherein: The large language model is fine-tuned on the Multi-Type Natural Language Inference (MNLI) dataset.

4. The computer-implemented method of claim 1 , further comprising: One or more predictions associated with the text are generated.

5. The computer-implemented method of claim 4, wherein: The one or more predictions are associated with response to a treatment administered for treatment-resistant depression.

6. The computer-implemented method of claim 1 , wherein: The sentiment score includes an arousal score, a valence score, and a confidence score for each portion of the text.

7. A computing system, comprising: computing device processor; and A memory device, the memory device comprising instructions, which when executed by a processor of the computing device, enable the computing system to perform the following operations: Receive text; Using a machine learning model to determine a set of classification probabilities for the text relative to a set of anchor points; as well as A sentiment score indicative of emotional content of the text is determined using the set of probabilities for the text based at least in part on a convex combination of the set of anchor points.

8. The computing system of claim 7, wherein: The classification probabilities are calculated using the output of a large language model.

9. The computing system of claim 8, wherein: The large language model is fine-tuned on the Multi-Type Natural Language Inference (MNLI) dataset.

10. The computing system of claim 7, wherein: When the instructions are executed by the computing device, the computing system is enabled to further perform the following operations: One or more predictions associated with the text are generated.

11. The computing system of claim 7, wherein: The sentiment score includes an arousal score, a valence score, and a confidence score for each portion of the text.

12. The computing system of claim 10, wherein: The one or more predictions are associated with response to a treatment administered for treatment-resistant depression.

13. The computing system of claim 7, wherein: When the instructions are executed by the computing device processor, the computing system is enabled to further perform the following operations: passing the text through a classifier; calculating a probability value indicating a probability that a text string of the text is located in a category of the one or more categories; dividing the one or more categories into a set of lists corresponding to the set of anchor points; and Weights are generated for respective character strings in the text based at least in part on the set of lists.

14. A non-transitory computer-readable medium storing instructions, which, when executed by at least one processor, cause the at least one processor to perform the following operations: Receive text; Using a machine learning model to determine a set of classification probabilities for the text relative to a set of anchor points; and A sentiment score indicative of emotional content of the text is determined using the set of probabilities for the text based at least in part on a convex combination of the anchor points.

15. The non-transitory computer readable medium of claim 14, wherein: The sentiment score is determined using a classifier built from a large language model.

16. The non-transitory computer readable medium of claim 15, wherein: The large language model is fine-tuned on the Multi-Type Natural Language Inference (MNLI) dataset.

17. The non-transitory computer readable medium of claim 14, wherein: When the instructions are executed by the at least one processor, the at least one processor is enabled to further perform the following operations: One or more predictions associated with the text are generated.

18. The non-transitory computer readable medium of claim 14, wherein: The sentiment score includes an arousal score, a valence score, and a confidence score for each portion of the text.

19. The non-transitory computer readable medium of claim 17, wherein: The one or more predictions are associated with response to a treatment administered for treatment-resistant depression.

20. The non-transitory computer readable medium of claim 14, wherein: When the instructions are executed by the at least one processor, the at least one processor is enabled to further perform the following operations: passing the text through a classifier; calculating a probability value indicating a probability that a text string in the set of texts is located in a category of the one or more categories; dividing the one or more categories into a set of lists corresponding to the set of anchor points; and Weights are generated for respective character strings in the text based at least in part on the set of lists.

Images